first commit

.gitignore

@@ -0,0 +1,30 @@
+# .gitignore
+# 首先忽略所有的文件
+*
+# 但是不忽略目录
+!*/
+# 忽略一些指定的目录名
+ut/
+runs/
+.vscode/
+build/
+result1/
+mytest/
+*.pyc
+# 不忽略下面指定的文件类型
+!*.cpp
+!*.h
+!*.hpp
+!*.c
+!.gitignore
+!*.py
+!*.sh
+!*.npy
+!*.jpg
+!*.pt
+!*.npy
+!*.pth
+!*.png
+!*.yaml
+!*.ttf
+!*.txt

Car_recognition.py

@@ -0,0 +1,350 @@
+# -*- coding: UTF-8 -*-
+import argparse
+import time
+import os
+import cv2
+import torch
+from numpy import random
+import copy
+import numpy as np
+from models.experimental import attempt_load
+from utils.datasets import letterbox
+from utils.general import check_img_size, non_max_suppression_face, scale_coords
+from utils.torch_utils import time_synchronized
+from utils.cv_puttext import cv2ImgAddText
+from plate_recognition.plate_rec import get_plate_result,allFilePath,init_model,cv_imread
+# from plate_recognition.plate_cls import cv_imread
+from plate_recognition.double_plate_split_merge import get_split_merge
+from plate_recognition.color_rec import plate_color_rec,init_color_model
+
+clors = [(255,0,0),(0,255,0),(0,0,255),(255,255,0),(0,255,255)]
+danger=['危','险']
+object_color=[(0,255,255),(0,255,0),(255,255,0)]
+class_type=['单层车牌','双层车牌','汽车']
+def order_points(pts):                   #四个点安好左上 右上 右下 左下排列
+    rect = np.zeros((4, 2), dtype = "float32")
+    s = pts.sum(axis = 1)
+    rect[0] = pts[np.argmin(s)]
+    rect[2] = pts[np.argmax(s)]
+    diff = np.diff(pts, axis = 1)
+    rect[1] = pts[np.argmin(diff)]
+    rect[3] = pts[np.argmax(diff)]
+    return rect
+
+
+def four_point_transform(image, pts):                       #透视变换得到车牌小图
+    rect = order_points(pts)
+    (tl, tr, br, bl) = rect
+    widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
+    widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
+    maxWidth = max(int(widthA), int(widthB))
+    heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
+    heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
+    maxHeight = max(int(heightA), int(heightB))
+    dst = np.array([
+        [0, 0],
+        [maxWidth - 1, 0],
+        [maxWidth - 1, maxHeight - 1],
+        [0, maxHeight - 1]], dtype = "float32")
+    M = cv2.getPerspectiveTransform(rect, dst)
+    warped = cv2.warpPerspective(image, M, (maxWidth, maxHeight))
+    return warped
+
+def load_model(weights, device):
+    model = attempt_load(weights, map_location=device)  # load FP32 model
+    return model
+
+def scale_coords_landmarks(img1_shape, coords, img0_shape, ratio_pad=None):  #返回到原图坐标
+    # Rescale coords (xyxy) from img1_shape to img0_shape
+    if ratio_pad is None:  # calculate from img0_shape
+        gain = min(img1_shape[0] / img0_shape[0], img1_shape[1] / img0_shape[1])  # gain  = old / new
+        pad = (img1_shape[1] - img0_shape[1] * gain) / 2, (img1_shape[0] - img0_shape[0] * gain) / 2  # wh padding
+    else:
+        gain = ratio_pad[0][0]
+        pad = ratio_pad[1]
+
+    coords[:, [0, 2, 4, 6]] -= pad[0]  # x padding
+    coords[:, [1, 3, 5, 7]] -= pad[1]  # y padding
+    coords[:, :10] /= gain
+    #clip_coords(coords, img0_shape)
+    coords[:, 0].clamp_(0, img0_shape[1])  # x1
+    coords[:, 1].clamp_(0, img0_shape[0])  # y1
+    coords[:, 2].clamp_(0, img0_shape[1])  # x2
+    coords[:, 3].clamp_(0, img0_shape[0])  # y2
+    coords[:, 4].clamp_(0, img0_shape[1])  # x3
+    coords[:, 5].clamp_(0, img0_shape[0])  # y3
+    coords[:, 6].clamp_(0, img0_shape[1])  # x4
+    coords[:, 7].clamp_(0, img0_shape[0])  # y4
+    # coords[:, 8].clamp_(0, img0_shape[1])  # x5
+    # coords[:, 9].clamp_(0, img0_shape[0])  # y5
+    return coords
+
+def get_plate_rec_landmark(img, xyxy, conf, landmarks, class_num,device,plate_rec_model,plate_color_model=None):
+    h,w,c = img.shape
+    result_dict={}
+    x1 = int(xyxy[0])
+    y1 = int(xyxy[1])
+    x2 = int(xyxy[2])
+    y2 = int(xyxy[3])
+    landmarks_np=np.zeros((4,2))
+    rect=[x1,y1,x2,y2]
+    
+    if int(class_num) ==2:
+        result_dict['class_type']=class_type[int(class_num)]
+        result_dict['rect']=rect                      #车辆roi
+        result_dict['score']=conf
+        result_dict['object_no']=int(class_num)
+        return result_dict
+    
+    for i in range(4):
+        point_x = int(landmarks[2 * i])
+        point_y = int(landmarks[2 * i + 1])
+        landmarks_np[i]=np.array([point_x,point_y])
+
+    class_label= int(class_num)  #车牌的的类型0代表单牌，1代表双层车牌
+    roi_img = four_point_transform(img,landmarks_np)   #透视变换得到车牌小图
+    color_code = plate_color_rec(roi_img,plate_color_model,device) #车牌颜色识别
+    if class_label:        #判断是否是双层车牌，是双牌的话进行分割后然后拼接
+        roi_img=get_split_merge(roi_img)
+    plate_number = get_plate_result(roi_img,device,plate_rec_model)                 #对车牌小图进行识别
+    for dan in danger:                                                           #只要出现‘危’或者‘险’就是危险品车牌
+        if dan in plate_number:
+            plate_number='危险品'
+    # cv2.imwrite("roi.jpg",roi_img)
+    result_dict['class_type']=class_type[class_label]
+    result_dict['rect']=rect                      #车牌roi区域
+    result_dict['landmarks']=landmarks_np.tolist() #车牌角点坐标
+    result_dict['plate_no']=plate_number   #车牌号
+    result_dict['roi_height']=roi_img.shape[0]  #车牌高度
+    result_dict['plate_color']=color_code   #车牌颜色
+    result_dict['object_no']=class_label   #单双层 0单层 1双层
+    result_dict['score']=conf
+    return result_dict
+
+
+
+def detect_Recognition_plate(model, orgimg, device,plate_rec_model,img_size,plate_color_model=None):
+    # Load model
+    # img_size = opt_img_size
+    conf_thres = 0.3
+    iou_thres = 0.5
+    dict_list=[]
+    # orgimg = cv2.imread(image_path)  # BGR
+    img0 = copy.deepcopy(orgimg)
+    assert orgimg is not None, 'Image Not Found ' 
+    h0, w0 = orgimg.shape[:2]  # orig hw
+    r = img_size / max(h0, w0)  # resize image to img_size
+    if r != 1:  # always resize down, only resize up if training with augmentation
+        interp = cv2.INTER_AREA if r < 1  else cv2.INTER_LINEAR
+        img0 = cv2.resize(img0, (int(w0 * r), int(h0 * r)), interpolation=interp)
+
+    imgsz = check_img_size(img_size, s=model.stride.max())  # check img_size
+
+    img = letterbox(img0, new_shape=imgsz)[0]
+    # img =process_data(img0)
+    # Convert
+    img = img[:, :, ::-1].transpose(2, 0, 1).copy()  # BGR to RGB, to 3x416x416
+
+    # Run inference
+    t0 = time.time()
+
+    img = torch.from_numpy(img).to(device)
+    img = img.float()  # uint8 to fp16/32
+    img /= 255.0  # 0 - 255 to 0.0 - 1.0
+    if img.ndimension() == 3:
+        img = img.unsqueeze(0)
+
+    # Inference
+    t1 = time_synchronized()
+    pred = model(img)[0]
+    t2=time_synchronized()
+    # print(f"infer time is {(t2-t1)*1000} ms")
+
+    # Apply NMS
+    pred = non_max_suppression_face(pred, conf_thres, iou_thres)
+
+    # print('img.shape: ', img.shape)
+    # print('orgimg.shape: ', orgimg.shape)
+
+    # Process detections
+    for i, det in enumerate(pred):  # detections per image
+        if len(det):
+            # Rescale boxes from img_size to im0 size
+            det[:, :4] = scale_coords(img.shape[2:], det[:, :4], orgimg.shape).round()
+
+            # Print results
+            for c in det[:, -1].unique():
+                n = (det[:, -1] == c).sum()  # detections per class
+
+            det[:, 5:13] = scale_coords_landmarks(img.shape[2:], det[:, 5:13], orgimg.shape).round()
+
+            for j in range(det.size()[0]):
+                xyxy = det[j, :4].view(-1).tolist()
+                conf = det[j, 4].cpu().numpy()
+                landmarks = det[j, 5:13].view(-1).tolist()
+                class_num = det[j, 13].cpu().numpy()
+                result_dict = get_plate_rec_landmark(orgimg, xyxy, conf, landmarks, class_num,device,plate_rec_model,plate_color_model)
+                dict_list.append(result_dict)
+    return dict_list
+    # cv2.imwrite('result.jpg', orgimg)
+
+def draw_result(orgimg,dict_list):
+    result_str =""
+    for result in dict_list:
+        rect_area = result['rect']
+        object_no = result['object_no']
+        if not object_no==2:    
+            x,y,w,h = rect_area[0],rect_area[1],rect_area[2]-rect_area[0],rect_area[3]-rect_area[1]
+            padding_w = 0.05*w
+            padding_h = 0.11*h
+            rect_area[0]=max(0,int(x-padding_w))
+            rect_area[1]=max(0,int(y-padding_h))
+            rect_area[2]=min(orgimg.shape[1],int(rect_area[2]+padding_w))
+            rect_area[3]=min(orgimg.shape[0],int(rect_area[3]+padding_h))
+
+            height_area = result['roi_height']
+            landmarks=result['landmarks']
+            result_p = result['plate_no']
+            if result['object_no']==0:#单层
+                result_p+=" "+result['plate_color']
+            else:                             #双层
+                result_p+=" "+result['plate_color']+"双层"
+            result_str+=result_p+" "
+            for i in range(4):  #关键点
+                cv2.circle(orgimg, (int(landmarks[i][0]), int(landmarks[i][1])), 5, clors[i], -1)
+            
+            if len(result)>=1:
+                if "危险品" in result_p: #如果是危险品车牌，文字就画在下面
+                    orgimg=cv2ImgAddText(orgimg,result_p,rect_area[0],rect_area[3],(0,255,0),height_area)
+                else:
+                    orgimg=cv2ImgAddText(orgimg,result_p,rect_area[0]-height_area,rect_area[1]-height_area-10,(0,255,0),height_area)
+        cv2.rectangle(orgimg,(rect_area[0],rect_area[1]),(rect_area[2],rect_area[3]),object_color[object_no],2) #画框       
+    print(result_str)
+    return orgimg
+
+def get_second(capture):
+    if capture.isOpened():
+        rate = capture.get(5)   # 帧速率
+        FrameNumber = capture.get(7)  # 视频文件的帧数
+        duration = FrameNumber/rate  # 帧速率/视频总帧数 是时间，除以60之后单位是分钟
+        return int(rate),int(FrameNumber),int(duration)    
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--detect_model', nargs='+', type=str, default='weights/detect.pt', help='model.pt path(s)')  #检测模型
+    parser.add_argument('--rec_model', type=str, default='weights/plate_rec.pth', help='model.pt path(s)')#识别模型
+    parser.add_argument('--color_model',type=str,default='weights/color_classify.pth',help='plate color')#颜色识别模型
+    parser.add_argument('--image_path', type=str, default='imgs', help='source') 
+    parser.add_argument('--img_size', type=int, default=384, help='inference size (pixels)')
+    parser.add_argument('--output', type=str, default='result1', help='source') 
+    parser.add_argument('--video', type=str, default='', help='source') 
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+    # device =torch.device("cpu")
+    opt = parser.parse_args()
+    print(opt)
+    save_path = opt.output
+    count=0
+    if not os.path.exists(save_path):
+        os.mkdir(save_path)
+
+    detect_model = load_model(opt.detect_model, device)  #初始化检测模型
+    plate_rec_model=init_model(device,opt.rec_model)      #初始化识别模型
+    #算参数量
+    total = sum(p.numel() for p in detect_model.parameters())
+    total_1 = sum(p.numel() for p in plate_rec_model.parameters())
+    print("detect params: %.2fM,rec params: %.2fM" % (total/1e6,total_1/1e6))
+    
+    plate_color_model =init_color_model(opt.color_model,device)
+    time_all = 0
+    time_begin=time.time()
+    if not opt.video:     #处理图片
+        if not os.path.isfile(opt.image_path):            #目录
+            file_list=[]
+            allFilePath(opt.image_path,file_list)
+            for img_path in file_list:
+                
+                print(count,img_path,end=" ")
+                time_b = time.time()
+                img =cv_imread(img_path)
+                
+                if img is None:
+                    continue
+                if img.shape[-1]==4:
+                    img=cv2.cvtColor(img,cv2.COLOR_BGRA2BGR)
+                # detect_one(model,img_path,device)
+                dict_list=detect_Recognition_plate(detect_model, img, device,plate_rec_model,opt.img_size,plate_color_model)
+                ori_img=draw_result(img,dict_list)
+                img_name = os.path.basename(img_path)
+                save_img_path = os.path.join(save_path,img_name)
+                time_e=time.time()
+                time_gap = time_e-time_b
+                if count:
+                    time_all+=time_gap
+                cv2.imwrite(save_img_path,ori_img)
+                count+=1
+        else:                                          #单个图片
+                print(count,opt.image_path,end=" ")
+                img =cv_imread(opt.image_path)
+                if img.shape[-1]==4:
+                    img=cv2.cvtColor(img,cv2.COLOR_BGRA2BGR)
+                # detect_one(model,img_path,device)
+                dict_list=detect_Recognition_plate(detect_model, img, device,plate_rec_model,opt.img_size)
+                ori_img=draw_result(img,dict_list)
+                img_name = os.path.basename(opt.image_path)
+                save_img_path = os.path.join(save_path,img_name)
+                cv2.imwrite(save_img_path,ori_img)  
+        print(f"sumTime time is {time.time()-time_begin} s, average pic time is {time_all/(len(file_list)-1)}")
+        
+    else:    #处理视频
+        video_name = opt.video
+        capture=cv2.VideoCapture(video_name)
+        fourcc = cv2.VideoWriter_fourcc(*'MP4V') 
+        fps = capture.get(cv2.CAP_PROP_FPS)  # 帧数
+        width, height = int(capture.get(cv2.CAP_PROP_FRAME_WIDTH)), int(capture.get(cv2.CAP_PROP_FRAME_HEIGHT))  # 宽高
+        out = cv2.VideoWriter('result.mp4', fourcc, fps, (width, height))  # 写入视频
+        frame_count = 0
+        fps_all=0
+        rate,FrameNumber,duration=get_second(capture)
+        if capture.isOpened():
+            while True:
+                t1 = cv2.getTickCount()
+                frame_count+=1
+                print(f"第{frame_count} 帧",end=" ")
+                ret,img=capture.read()
+                if not ret:
+                    break
+                # if frame_count%rate==0:
+                img0 = copy.deepcopy(img)
+                dict_list=detect_Recognition_plate(detect_model, img, device,plate_rec_model,opt.img_size,plate_color_model)
+                ori_img=draw_result(img,dict_list)
+                t2 =cv2.getTickCount()
+                infer_time =(t2-t1)/cv2.getTickFrequency()
+                fps=1.0/infer_time
+                fps_all+=fps
+                str_fps = f'fps:{fps:.4f}'
+                
+                cv2.putText(ori_img,str_fps,(20,20),cv2.FONT_HERSHEY_SIMPLEX,1,(0,255,0),2)
+                # cv2.imshow("haha",ori_img)
+                # cv2.waitKey(1)
+                out.write(ori_img)
+
+                # current_time = int(frame_count/FrameNumber*duration)
+                # sec = current_time%60
+                # minute = current_time//60
+                # for result_ in result_list:
+                #     plate_no = result_['plate_no']
+                #     if not is_car_number(pattern_str,plate_no):
+                #         continue
+                #     print(f'车牌号:{plate_no},时间:{minute}分{sec}秒')
+                #     time_str =f'{minute}分{sec}秒'
+                #     writer.writerow({"车牌":plate_no,"时间":time_str})
+                # out.write(ori_img)
+                
+                
+        else:
+            print("失败")
+        capture.release()
+        out.release()
+        cv2.destroyAllWindows()
+        print(f"all frame is {frame_count},average fps is {fps_all/frame_count} fps")

data/argoverse_hd.yaml

@@ -0,0 +1,21 @@
+# Argoverse-HD dataset (ring-front-center camera) http://www.cs.cmu.edu/~mengtial/proj/streaming/
+# Train command: python train.py --data argoverse_hd.yaml
+# Default dataset location is next to /yolov5:
+#   /parent_folder
+#     /argoverse
+#     /yolov5
+
+
+# download command/URL (optional)
+download: bash data/scripts/get_argoverse_hd.sh
+
+# train and val data as 1) directory: path/images/, 2) file: path/images.txt, or 3) list: [path1/images/, path2/images/]
+train: ../argoverse/Argoverse-1.1/images/train/  # 39384 images
+val: ../argoverse/Argoverse-1.1/images/val/  # 15062 iamges
+test: ../argoverse/Argoverse-1.1/images/test/  # Submit to: https://eval.ai/web/challenges/challenge-page/800/overview
+
+# number of classes
+nc: 8
+
+# class names
+names: [ 'person',  'bicycle',  'car',  'motorcycle',  'bus',  'truck',  'traffic_light',  'stop_sign' ]

data/coco.yaml

@@ -0,0 +1,35 @@
+# COCO 2017 dataset http://cocodataset.org
+# Train command: python train.py --data coco.yaml
+# Default dataset location is next to /yolov5:
+#   /parent_folder
+#     /coco
+#     /yolov5
+
+
+# download command/URL (optional)
+download: bash data/scripts/get_coco.sh
+
+# train and val data as 1) directory: path/images/, 2) file: path/images.txt, or 3) list: [path1/images/, path2/images/]
+train: ../coco/train2017.txt  # 118287 images
+val: ../coco/val2017.txt  # 5000 images
+test: ../coco/test-dev2017.txt  # 20288 of 40670 images, submit to https://competitions.codalab.org/competitions/20794
+
+# number of classes
+nc: 80
+
+# class names
+names: [ 'person', 'bicycle', 'car', 'motorcycle', 'airplane', 'bus', 'train', 'truck', 'boat', 'traffic light',
+         'fire hydrant', 'stop sign', 'parking meter', 'bench', 'bird', 'cat', 'dog', 'horse', 'sheep', 'cow',
+         'elephant', 'bear', 'zebra', 'giraffe', 'backpack', 'umbrella', 'handbag', 'tie', 'suitcase', 'frisbee',
+         'skis', 'snowboard', 'sports ball', 'kite', 'baseball bat', 'baseball glove', 'skateboard', 'surfboard',
+         'tennis racket', 'bottle', 'wine glass', 'cup', 'fork', 'knife', 'spoon', 'bowl', 'banana', 'apple',
+         'sandwich', 'orange', 'broccoli', 'carrot', 'hot dog', 'pizza', 'donut', 'cake', 'chair', 'couch',
+         'potted plant', 'bed', 'dining table', 'toilet', 'tv', 'laptop', 'mouse', 'remote', 'keyboard', 'cell phone',
+         'microwave', 'oven', 'toaster', 'sink', 'refrigerator', 'book', 'clock', 'vase', 'scissors', 'teddy bear',
+         'hair drier', 'toothbrush' ]
+
+# Print classes
+# with open('data/coco.yaml') as f:
+#   d = yaml.load(f, Loader=yaml.FullLoader)  # dict
+#   for i, x in enumerate(d['names']):
+#     print(i, x)

data/coco128.yaml

@@ -0,0 +1,28 @@
+# COCO 2017 dataset http://cocodataset.org - first 128 training images
+# Train command: python train.py --data coco128.yaml
+# Default dataset location is next to /yolov5:
+#   /parent_folder
+#     /coco128
+#     /yolov5
+
+
+# download command/URL (optional)
+download: https://github.com/ultralytics/yolov5/releases/download/v1.0/coco128.zip
+
+# train and val data as 1) directory: path/images/, 2) file: path/images.txt, or 3) list: [path1/images/, path2/images/]
+train: ../coco128/images/train2017/  # 128 images
+val: ../coco128/images/train2017/  # 128 images
+
+# number of classes
+nc: 80
+
+# class names
+names: [ 'person', 'bicycle', 'car', 'motorcycle', 'airplane', 'bus', 'train', 'truck', 'boat', 'traffic light',
+         'fire hydrant', 'stop sign', 'parking meter', 'bench', 'bird', 'cat', 'dog', 'horse', 'sheep', 'cow',
+         'elephant', 'bear', 'zebra', 'giraffe', 'backpack', 'umbrella', 'handbag', 'tie', 'suitcase', 'frisbee',
+         'skis', 'snowboard', 'sports ball', 'kite', 'baseball bat', 'baseball glove', 'skateboard', 'surfboard',
+         'tennis racket', 'bottle', 'wine glass', 'cup', 'fork', 'knife', 'spoon', 'bowl', 'banana', 'apple',
+         'sandwich', 'orange', 'broccoli', 'carrot', 'hot dog', 'pizza', 'donut', 'cake', 'chair', 'couch',
+         'potted plant', 'bed', 'dining table', 'toilet', 'tv', 'laptop', 'mouse', 'remote', 'keyboard', 'cell phone',
+         'microwave', 'oven', 'toaster', 'sink', 'refrigerator', 'book', 'clock', 'vase', 'scissors', 'teddy bear',
+         'hair drier', 'toothbrush' ]

data/hyp.finetune.yaml

@@ -0,0 +1,38 @@
+# Hyperparameters for VOC finetuning
+# python train.py --batch 64 --weights yolov5m.pt --data voc.yaml --img 512 --epochs 50
+# See tutorials for hyperparameter evolution https://github.com/ultralytics/yolov5#tutorials
+
+
+# Hyperparameter Evolution Results
+# Generations: 306
+#                   P         R     mAP.5 mAP.5:.95       box       obj       cls
+# Metrics:        0.6     0.936     0.896     0.684    0.0115   0.00805   0.00146
+
+lr0: 0.0032
+lrf: 0.12
+momentum: 0.843
+weight_decay: 0.00036
+warmup_epochs: 2.0
+warmup_momentum: 0.5
+warmup_bias_lr: 0.05
+box: 0.0296
+cls: 0.243
+cls_pw: 0.631
+obj: 0.301
+obj_pw: 0.911
+iou_t: 0.2
+anchor_t: 2.91
+# anchors: 3.63
+fl_gamma: 0.0
+hsv_h: 0.0138
+hsv_s: 0.664
+hsv_v: 0.464
+degrees: 0.373
+translate: 0.245
+scale: 0.898
+shear: 0.602
+perspective: 0.0
+flipud: 0.00856
+fliplr: 0.5
+mosaic: 1.0
+mixup: 0.243

data/hyp.scratch.yaml

@@ -0,0 +1,34 @@
+# Hyperparameters for COCO training from scratch
+# python train.py --batch 40 --cfg yolov5m.yaml --weights '' --data coco.yaml --img 640 --epochs 300
+# See tutorials for hyperparameter evolution https://github.com/ultralytics/yolov5#tutorials
+
+
+lr0: 0.01  # initial learning rate (SGD=1E-2, Adam=1E-3)
+lrf: 0.2  # final OneCycleLR learning rate (lr0 * lrf)
+momentum: 0.937  # SGD momentum/Adam beta1
+weight_decay: 0.0005  # optimizer weight decay 5e-4
+warmup_epochs: 3.0  # warmup epochs (fractions ok)
+warmup_momentum: 0.8  # warmup initial momentum
+warmup_bias_lr: 0.1  # warmup initial bias lr
+box: 0.05  # box loss gain
+cls: 0.5  # cls loss gain
+landmark: 0.005 # landmark loss gain
+cls_pw: 1.0  # cls BCELoss positive_weight
+obj: 1.0  # obj loss gain (scale with pixels)
+obj_pw: 1.0  # obj BCELoss positive_weight
+iou_t: 0.20  # IoU training threshold
+anchor_t: 4.0  # anchor-multiple threshold
+# anchors: 3  # anchors per output layer (0 to ignore)
+fl_gamma: 0.0  # focal loss gamma (efficientDet default gamma=1.5)
+hsv_h: 0.015  # image HSV-Hue augmentation (fraction)
+hsv_s: 0.7  # image HSV-Saturation augmentation (fraction)
+hsv_v: 0.4  # image HSV-Value augmentation (fraction)
+degrees: 0.0  # image rotation (+/- deg)
+translate: 0.1  # image translation (+/- fraction)
+scale: 0.5  # image scale (+/- gain)
+shear: 0.5  # image shear (+/- deg)
+perspective: 0.0  # image perspective (+/- fraction), range 0-0.001
+flipud: 0.0  # image flip up-down (probability)
+fliplr: 0.5  # image flip left-right (probability)
+mosaic: 0.5  # image mosaic (probability)
+mixup: 0.0  # image mixup (probability)

data/plateAndCar.yaml

@@ -0,0 +1,19 @@
+# PASCAL VOC dataset http://host.robots.ox.ac.uk/pascal/VOC/
+# Train command: python train.py --data voc.yaml
+# Default dataset location is next to /yolov5:
+#   /parent_folder
+#     /VOC
+#     /yolov5
+
+
+# download command/URL (optional)
+download: bash data/scripts/get_voc.sh
+
+# train and val data as 1) directory: path/images/, 2) file: path/images.txt, or 3) list: [path1/images/, path2/images/]
+train: /mnt/Gpan/Mydata/pytorchPorject/datasets/ccpd/train_car_plate/train_detect  
+val: /mnt/Gpan/Mydata/pytorchPorject/datasets/ccpd/train_car_plate/val_detect
+# number of classes
+nc: 3
+
+# class names
+names: [ 'single_plate','double_plate','car']

data/retinaface2yolo.py

@@ -0,0 +1,150 @@
+import os
+import os.path
+import sys
+import torch
+import torch.utils.data as data
+import cv2
+import numpy as np
+
+class WiderFaceDetection(data.Dataset):
+    def __init__(self, txt_path, preproc=None):
+        self.preproc = preproc
+        self.imgs_path = []
+        self.words = []
+        f = open(txt_path,'r')
+        lines = f.readlines()
+        isFirst = True
+        labels = []
+        for line in lines:
+            line = line.rstrip()
+            if line.startswith('#'):
+                if isFirst is True:
+                    isFirst = False
+                else:
+                    labels_copy = labels.copy()
+                    self.words.append(labels_copy)
+                    labels.clear()
+                path = line[2:]
+                path = txt_path.replace('label.txt','images/') + path
+                self.imgs_path.append(path)
+            else:
+                line = line.split(' ')
+                label = [float(x) for x in line]
+                labels.append(label)
+
+        self.words.append(labels)
+
+    def __len__(self):
+        return len(self.imgs_path)
+
+    def __getitem__(self, index):
+        img = cv2.imread(self.imgs_path[index])
+        height, width, _ = img.shape
+
+        labels = self.words[index]
+        annotations = np.zeros((0, 15))
+        if len(labels) == 0:
+            return annotations
+        for idx, label in enumerate(labels):
+            annotation = np.zeros((1, 15))
+            # bbox
+            annotation[0, 0] = label[0]  # x1
+            annotation[0, 1] = label[1]  # y1
+            annotation[0, 2] = label[0] + label[2]  # x2
+            annotation[0, 3] = label[1] + label[3]  # y2
+
+            # landmarks
+            annotation[0, 4] = label[4]    # l0_x
+            annotation[0, 5] = label[5]    # l0_y
+            annotation[0, 6] = label[7]    # l1_x
+            annotation[0, 7] = label[8]    # l1_y
+            annotation[0, 8] = label[10]   # l2_x
+            annotation[0, 9] = label[11]   # l2_y
+            annotation[0, 10] = label[13]  # l3_x
+            annotation[0, 11] = label[14]  # l3_y
+            annotation[0, 12] = label[16]  # l4_x
+            annotation[0, 13] = label[17]  # l4_y
+            if (annotation[0, 4]<0):
+                annotation[0, 14] = -1
+            else:
+                annotation[0, 14] = 1
+
+            annotations = np.append(annotations, annotation, axis=0)
+        target = np.array(annotations)
+        if self.preproc is not None:
+            img, target = self.preproc(img, target)
+
+        return torch.from_numpy(img), target
+
+def detection_collate(batch):
+    """Custom collate fn for dealing with batches of images that have a different
+    number of associated object annotations (bounding boxes).
+
+    Arguments:
+        batch: (tuple) A tuple of tensor images and lists of annotations
+
+    Return:
+        A tuple containing:
+            1) (tensor) batch of images stacked on their 0 dim
+            2) (list of tensors) annotations for a given image are stacked on 0 dim
+    """
+    targets = []
+    imgs = []
+    for _, sample in enumerate(batch):
+        for _, tup in enumerate(sample):
+            if torch.is_tensor(tup):
+                imgs.append(tup)
+            elif isinstance(tup, type(np.empty(0))):
+                annos = torch.from_numpy(tup).float()
+                targets.append(annos)
+
+    return (torch.stack(imgs, 0), targets)
+
+save_path = '/ssd_1t/derron/yolov5-face/data/widerface/train'
+aa=WiderFaceDetection("/ssd_1t/derron/yolov5-face/data/widerface/widerface/train/label.txt")
+for i in range(len(aa.imgs_path)):
+    print(i, aa.imgs_path[i])
+    img = cv2.imread(aa.imgs_path[i])
+    base_img = os.path.basename(aa.imgs_path[i])
+    base_txt = os.path.basename(aa.imgs_path[i])[:-4] +".txt"
+    save_img_path = os.path.join(save_path, base_img)
+    save_txt_path = os.path.join(save_path, base_txt)
+    with open(save_txt_path, "w") as f:
+        height, width, _ = img.shape
+        labels = aa.words[i]
+        annotations = np.zeros((0, 14))
+        if len(labels) == 0:
+            continue
+        for idx, label in enumerate(labels):
+            annotation = np.zeros((1, 14))
+            # bbox
+            label[0] = max(0, label[0])
+            label[1] = max(0, label[1])
+            label[2] = min(width -  1, label[2])
+            label[3] = min(height - 1, label[3])
+            annotation[0, 0] = (label[0] + label[2] / 2) / width  # cx
+            annotation[0, 1] = (label[1] + label[3] / 2) / height  # cy
+            annotation[0, 2] = label[2] / width  # w
+            annotation[0, 3] = label[3] / height  # h
+            #if (label[2] -label[0]) < 8 or (label[3] - label[1]) < 8:
+            #    img[int(label[1]):int(label[3]), int(label[0]):int(label[2])] = 127
+            #    continue
+            # landmarks
+            annotation[0, 4] = label[4] / width  # l0_x
+            annotation[0, 5] = label[5] / height  # l0_y
+            annotation[0, 6] = label[7] / width  # l1_x
+            annotation[0, 7] = label[8]  / height # l1_y
+            annotation[0, 8] = label[10] / width  # l2_x
+            annotation[0, 9] = label[11] / height  # l2_y
+            annotation[0, 10] = label[13] / width  # l3_x
+            annotation[0, 11] = label[14] / height  # l3_y
+            annotation[0, 12] = label[16] / width  # l4_x
+            annotation[0, 13] = label[17] / height  # l4_y
+            str_label="0 "
+            for i in range(len(annotation[0])):
+                str_label =str_label+" "+str(annotation[0][i])
+            str_label = str_label.replace('[', '').replace(']', '')
+            str_label = str_label.replace(',', '') + '\n'
+            f.write(str_label)
+    cv2.imwrite(save_img_path, img)
+

data/scripts/get_argoverse_hd.sh

@@ -0,0 +1,62 @@
+#!/bin/bash
+# Argoverse-HD dataset (ring-front-center camera) http://www.cs.cmu.edu/~mengtial/proj/streaming/
+# Download command: bash data/scripts/get_argoverse_hd.sh
+# Train command: python train.py --data argoverse_hd.yaml
+# Default dataset location is next to /yolov5:
+#   /parent_folder
+#     /argoverse
+#     /yolov5
+
+# Download/unzip images
+d='../argoverse/' # unzip directory
+mkdir $d
+url=https://argoverse-hd.s3.us-east-2.amazonaws.com/
+f=Argoverse-HD-Full.zip
+curl -L $url$f -o $f && unzip -q $f -d $d && rm $f &# download, unzip, remove in background
+wait                                              # finish background tasks
+
+cd ../argoverse/Argoverse-1.1/
+ln -s tracking images
+
+cd ../Argoverse-HD/annotations/
+
+python3 - "$@" <<END
+import json
+from pathlib import Path
+
+annotation_files = ["train.json", "val.json"]
+print("Converting annotations to YOLOv5 format...")
+
+for val in annotation_files:
+    a = json.load(open(val, "rb"))
+
+    label_dict = {}
+    for annot in a['annotations']:
+        img_id = annot['image_id']
+        img_name = a['images'][img_id]['name']
+        img_label_name = img_name[:-3] + "txt"
+
+        obj_class = annot['category_id']
+        x_center, y_center, width, height = annot['bbox']
+        x_center = (x_center + width / 2) / 1920.  # offset and scale
+        y_center = (y_center + height / 2) / 1200.  # offset and scale
+        width /= 1920.  # scale
+        height /= 1200.  # scale
+
+        img_dir = "./labels/" + a['seq_dirs'][a['images'][annot['image_id']]['sid']]
+
+        Path(img_dir).mkdir(parents=True, exist_ok=True)
+
+        if img_dir + "/" + img_label_name not in label_dict:
+            label_dict[img_dir + "/" + img_label_name] = []
+
+        label_dict[img_dir + "/" + img_label_name].append(f"{obj_class} {x_center} {y_center} {width} {height}\n")
+
+    for filename in label_dict:
+        with open(filename, "w") as file:
+            for string in label_dict[filename]:
+                file.write(string)
+
+END
+
+mv ./labels ../../Argoverse-1.1/

data/scripts/get_coco.sh

@@ -0,0 +1,27 @@
+#!/bin/bash
+# COCO 2017 dataset http://cocodataset.org
+# Download command: bash data/scripts/get_coco.sh
+# Train command: python train.py --data coco.yaml
+# Default dataset location is next to /yolov5:
+#   /parent_folder
+#     /coco
+#     /yolov5
+
+# Download/unzip labels
+d='../' # unzip directory
+url=https://github.com/ultralytics/yolov5/releases/download/v1.0/
+f='coco2017labels.zip' # or 'coco2017labels-segments.zip', 68 MB
+echo 'Downloading' $url$f ' ...'
+curl -L $url$f -o $f && unzip -q $f -d $d && rm $f & # download, unzip, remove in background
+
+# Download/unzip images
+d='../coco/images' # unzip directory
+url=http://images.cocodataset.org/zips/
+f1='train2017.zip' # 19G, 118k images
+f2='val2017.zip'   # 1G, 5k images
+f3='test2017.zip'  # 7G, 41k images (optional)
+for f in $f1 $f2; do
+  echo 'Downloading' $url$f '...'
+  curl -L $url$f -o $f && unzip -q $f -d $d && rm $f & # download, unzip, remove in background
+done
+wait # finish background tasks

data/scripts/get_voc.sh

@@ -0,0 +1,139 @@
+#!/bin/bash
+# PASCAL VOC dataset http://host.robots.ox.ac.uk/pascal/VOC/
+# Download command: bash data/scripts/get_voc.sh
+# Train command: python train.py --data voc.yaml
+# Default dataset location is next to /yolov5:
+#   /parent_folder
+#     /VOC
+#     /yolov5
+
+start=$(date +%s)
+mkdir -p ../tmp
+cd ../tmp/
+
+# Download/unzip images and labels
+d='.' # unzip directory
+url=https://github.com/ultralytics/yolov5/releases/download/v1.0/
+f1=VOCtrainval_06-Nov-2007.zip # 446MB, 5012 images
+f2=VOCtest_06-Nov-2007.zip     # 438MB, 4953 images
+f3=VOCtrainval_11-May-2012.zip # 1.95GB, 17126 images
+for f in $f3 $f2 $f1; do
+  echo 'Downloading' $url$f '...' 
+  curl -L $url$f -o $f && unzip -q $f -d $d && rm $f & # download, unzip, remove in background
+done
+wait # finish background tasks
+
+end=$(date +%s)
+runtime=$((end - start))
+echo "Completed in" $runtime "seconds"
+
+echo "Splitting dataset..."
+python3 - "$@" <<END
+import xml.etree.ElementTree as ET
+import pickle
+import os
+from os import listdir, getcwd
+from os.path import join
+
+sets=[('2012', 'train'), ('2012', 'val'), ('2007', 'train'), ('2007', 'val'), ('2007', 'test')]
+
+classes = ["aeroplane", "bicycle", "bird", "boat", "bottle", "bus", "car", "cat", "chair", "cow", "diningtable", "dog", "horse", "motorbike", "person", "pottedplant", "sheep", "sofa", "train", "tvmonitor"]
+
+
+def convert(size, box):
+    dw = 1./(size[0])
+    dh = 1./(size[1])
+    x = (box[0] + box[1])/2.0 - 1
+    y = (box[2] + box[3])/2.0 - 1
+    w = box[1] - box[0]
+    h = box[3] - box[2]
+    x = x*dw
+    w = w*dw
+    y = y*dh
+    h = h*dh
+    return (x,y,w,h)
+
+def convert_annotation(year, image_id):
+    in_file = open('VOCdevkit/VOC%s/Annotations/%s.xml'%(year, image_id))
+    out_file = open('VOCdevkit/VOC%s/labels/%s.txt'%(year, image_id), 'w')
+    tree=ET.parse(in_file)
+    root = tree.getroot()
+    size = root.find('size')
+    w = int(size.find('width').text)
+    h = int(size.find('height').text)
+
+    for obj in root.iter('object'):
+        difficult = obj.find('difficult').text
+        cls = obj.find('name').text
+        if cls not in classes or int(difficult)==1:
+            continue
+        cls_id = classes.index(cls)
+        xmlbox = obj.find('bndbox')
+        b = (float(xmlbox.find('xmin').text), float(xmlbox.find('xmax').text), float(xmlbox.find('ymin').text), float(xmlbox.find('ymax').text))
+        bb = convert((w,h), b)
+        out_file.write(str(cls_id) + " " + " ".join([str(a) for a in bb]) + '\n')
+
+wd = getcwd()
+
+for year, image_set in sets:
+    if not os.path.exists('VOCdevkit/VOC%s/labels/'%(year)):
+        os.makedirs('VOCdevkit/VOC%s/labels/'%(year))
+    image_ids = open('VOCdevkit/VOC%s/ImageSets/Main/%s.txt'%(year, image_set)).read().strip().split()
+    list_file = open('%s_%s.txt'%(year, image_set), 'w')
+    for image_id in image_ids:
+        list_file.write('%s/VOCdevkit/VOC%s/JPEGImages/%s.jpg\n'%(wd, year, image_id))
+        convert_annotation(year, image_id)
+    list_file.close()
+
+END
+
+cat 2007_train.txt 2007_val.txt 2012_train.txt 2012_val.txt >train.txt
+cat 2007_train.txt 2007_val.txt 2007_test.txt 2012_train.txt 2012_val.txt >train.all.txt
+
+python3 - "$@" <<END
+
+import shutil
+import os
+os.system('mkdir ../VOC/')
+os.system('mkdir ../VOC/images')
+os.system('mkdir ../VOC/images/train')
+os.system('mkdir ../VOC/images/val')
+
+os.system('mkdir ../VOC/labels')
+os.system('mkdir ../VOC/labels/train')
+os.system('mkdir ../VOC/labels/val')
+
+import os
+print(os.path.exists('../tmp/train.txt'))
+f = open('../tmp/train.txt', 'r')
+lines = f.readlines()
+
+for line in lines:
+    line = "/".join(line.split('/')[-5:]).strip()
+    if (os.path.exists("../" + line)):
+        os.system("cp ../"+ line + " ../VOC/images/train")
+        
+    line = line.replace('JPEGImages', 'labels')
+    line = line.replace('jpg', 'txt')
+    if (os.path.exists("../" + line)):
+        os.system("cp ../"+ line + " ../VOC/labels/train")
+
+
+print(os.path.exists('../tmp/2007_test.txt'))
+f = open('../tmp/2007_test.txt', 'r')
+lines = f.readlines()
+
+for line in lines:
+    line = "/".join(line.split('/')[-5:]).strip()
+    if (os.path.exists("../" + line)):
+        os.system("cp ../"+ line + " ../VOC/images/val")
+        
+    line = line.replace('JPEGImages', 'labels')
+    line = line.replace('jpg', 'txt')
+    if (os.path.exists("../" + line)):
+        os.system("cp ../"+ line + " ../VOC/labels/val")
+
+END
+
+rm -rf ../tmp # remove temporary directory
+echo "VOC download done."

data/train2yolo.py

@@ -0,0 +1,176 @@
+import os.path
+import sys
+import torch
+import torch.utils.data as data
+import cv2
+import numpy as np
+
+
+class WiderFaceDetection(data.Dataset):
+    def __init__(self, txt_path, preproc=None):
+        self.preproc = preproc
+        self.imgs_path = []
+        self.words = []
+        f = open(txt_path, 'r')
+        lines = f.readlines()
+        isFirst = True
+        labels = []
+        for line in lines:
+            line = line.rstrip()
+            if line.startswith('#'):
+                if isFirst is True:
+                    isFirst = False
+                else:
+                    labels_copy = labels.copy()
+                    self.words.append(labels_copy)
+                    labels.clear()
+                path = line[2:]
+                path = txt_path.replace('label.txt', 'images/') + path
+                self.imgs_path.append(path)
+            else:
+                line = line.split(' ')
+                label = [float(x) for x in line]
+                labels.append(label)
+
+        self.words.append(labels)
+
+    def __len__(self):
+        return len(self.imgs_path)
+
+    def __getitem__(self, index):
+        img = cv2.imread(self.imgs_path[index])
+        height, width, _ = img.shape
+
+        labels = self.words[index]
+        annotations = np.zeros((0, 15))
+        if len(labels) == 0:
+            return annotations
+        for idx, label in enumerate(labels):
+            annotation = np.zeros((1, 15))
+            # bbox
+            annotation[0, 0] = label[0]  # x1
+            annotation[0, 1] = label[1]  # y1
+            annotation[0, 2] = label[0] + label[2]  # x2
+            annotation[0, 3] = label[1] + label[3]  # y2
+
+            # landmarks
+            annotation[0, 4] = label[4]    # l0_x
+            annotation[0, 5] = label[5]    # l0_y
+            annotation[0, 6] = label[7]    # l1_x
+            annotation[0, 7] = label[8]    # l1_y
+            annotation[0, 8] = label[10]   # l2_x
+            annotation[0, 9] = label[11]   # l2_y
+            annotation[0, 10] = label[13]  # l3_x
+            annotation[0, 11] = label[14]  # l3_y
+            annotation[0, 12] = label[16]  # l4_x
+            annotation[0, 13] = label[17]  # l4_y
+            if annotation[0, 4] < 0:
+                annotation[0, 14] = -1
+            else:
+                annotation[0, 14] = 1
+
+            annotations = np.append(annotations, annotation, axis=0)
+        target = np.array(annotations)
+        if self.preproc is not None:
+            img, target = self.preproc(img, target)
+
+        return torch.from_numpy(img), target
+
+
+def detection_collate(batch):
+    """Custom collate fn for dealing with batches of images that have a different
+    number of associated object annotations (bounding boxes).
+
+    Arguments:
+        batch: (tuple) A tuple of tensor images and lists of annotations
+
+    Return:
+        A tuple containing:
+            1) (tensor) batch of images stacked on their 0 dim
+            2) (list of tensors) annotations for a given image are stacked on 0 dim
+    """
+    targets = []
+    imgs = []
+    for _, sample in enumerate(batch):
+        for _, tup in enumerate(sample):
+            if torch.is_tensor(tup):
+                imgs.append(tup)
+            elif isinstance(tup, type(np.empty(0))):
+                annos = torch.from_numpy(tup).float()
+                targets.append(annos)
+
+    return torch.stack(imgs, 0), targets
+
+
+if __name__ == '__main__':
+    if len(sys.argv) == 1:
+        print('Missing path to WIDERFACE train folder.')
+        print('Run command: python3 train2yolo.py /path/to/original/widerface/train [/path/to/save/widerface/train]')
+        exit(1)
+    elif len(sys.argv) > 3:
+        print('Too many arguments were provided.')
+        print('Run command: python3 train2yolo.py /path/to/original/widerface/train [/path/to/save/widerface/train]')
+        exit(1)
+    original_path = sys.argv[1]
+
+    if len(sys.argv) == 2:
+        if not os.path.isdir('widerface'):
+            os.mkdir('widerface')
+        if not os.path.isdir('widerface/train'):
+            os.mkdir('widerface/train')
+
+        save_path = 'widerface/train'
+    else:
+        save_path = sys.argv[2]
+
+    if not os.path.isfile(os.path.join(original_path, 'label.txt')):
+        print('Missing label.txt file.')
+        exit(1)
+
+    aa = WiderFaceDetection(os.path.join(original_path, 'label.txt'))
+
+    for i in range(len(aa.imgs_path)):
+        print(i, aa.imgs_path[i])
+        img = cv2.imread(aa.imgs_path[i])
+        base_img = os.path.basename(aa.imgs_path[i])
+        base_txt = os.path.basename(aa.imgs_path[i])[:-4] + ".txt"
+        save_img_path = os.path.join(save_path, base_img)
+        save_txt_path = os.path.join(save_path, base_txt)
+        with open(save_txt_path, "w") as f:
+            height, width, _ = img.shape
+            labels = aa.words[i]
+            annotations = np.zeros((0, 14))
+            if len(labels) == 0:
+                continue
+            for idx, label in enumerate(labels):
+                annotation = np.zeros((1, 14))
+                # bbox
+                label[0] = max(0, label[0])
+                label[1] = max(0, label[1])
+                label[2] = min(width - 1, label[2])
+                label[3] = min(height - 1, label[3])
+                annotation[0, 0] = (label[0] + label[2] / 2) / width  # cx
+                annotation[0, 1] = (label[1] + label[3] / 2) / height  # cy
+                annotation[0, 2] = label[2] / width  # w
+                annotation[0, 3] = label[3] / height  # h
+                #if (label[2] -label[0]) < 8 or (label[3] - label[1]) < 8:
+                #    img[int(label[1]):int(label[3]), int(label[0]):int(label[2])] = 127
+                #    continue
+                # landmarks
+                annotation[0, 4] = label[4] / width  # l0_x
+                annotation[0, 5] = label[5] / height  # l0_y
+                annotation[0, 6] = label[7] / width  # l1_x
+                annotation[0, 7] = label[8] / height  # l1_y
+                annotation[0, 8] = label[10] / width  # l2_x
+                annotation[0, 9] = label[11] / height  # l2_y
+                annotation[0, 10] = label[13] / width  # l3_x
+                annotation[0, 11] = label[14] / height  # l3_y
+                annotation[0, 12] = label[16] / width  # l4_x
+                annotation[0, 13] = label[17] / height  # l4_yca
+                str_label = "0 "
+                for i in range(len(annotation[0])):
+                    str_label = str_label + " " + str(annotation[0][i])
+                str_label = str_label.replace('[', '').replace(']', '')
+                str_label = str_label.replace(',', '') + '\n'
+                f.write(str_label)
+        cv2.imwrite(save_img_path, img)

data/val2yolo.py

@@ -0,0 +1,88 @@
+import os
+import cv2
+import numpy as np
+import shutil
+import sys
+from tqdm import tqdm
+
+
+def xywh2xxyy(box):
+    x1 = box[0]
+    y1 = box[1]
+    x2 = box[0] + box[2]
+    y2 = box[1] + box[3]
+    return x1, x2, y1, y2
+
+
+def convert(size, box):
+    dw = 1. / (size[0])
+    dh = 1. / (size[1])
+    x = (box[0] + box[1]) / 2.0 - 1
+    y = (box[2] + box[3]) / 2.0 - 1
+    w = box[1] - box[0]
+    h = box[3] - box[2]
+    x = x * dw
+    w = w * dw
+    y = y * dh
+    h = h * dh
+    return x, y, w, h
+
+
+def wider2face(root, phase='val', ignore_small=0):
+    data = {}
+    with open('{}/{}/label.txt'.format(root, phase), 'r') as f:
+        lines = f.readlines()
+        for line in tqdm(lines):
+            line = line.strip()
+            if '#' in line:
+                path = '{}/{}/images/{}'.format(root, phase, line.split()[-1])
+                img = cv2.imread(path)
+                height, width, _ = img.shape
+                data[path] = list()
+            else:
+                box = np.array(line.split()[0:4], dtype=np.float32)  # (x1,y1,w,h)
+                if box[2] < ignore_small or box[3] < ignore_small:
+                    continue
+                box = convert((width, height), xywh2xxyy(box))
+                label = '0 {} {} {} {} -1 -1 -1 -1 -1 -1 -1 -1 -1 -1'.format(round(box[0], 4), round(box[1], 4),
+                                                                             round(box[2], 4), round(box[3], 4))
+                data[path].append(label)
+    return data
+
+
+if __name__ == '__main__':
+    if len(sys.argv) == 1:
+        print('Missing path to WIDERFACE folder.')
+        print('Run command: python3 val2yolo.py /path/to/original/widerface [/path/to/save/widerface/val]')
+        exit(1)
+    elif len(sys.argv) > 3:
+        print('Too many arguments were provided.')
+        print('Run command: python3 val2yolo.py /path/to/original/widerface [/path/to/save/widerface/val]')
+        exit(1)
+
+    root_path = sys.argv[1]
+    if not os.path.isfile(os.path.join(root_path, 'val', 'label.txt')):
+        print('Missing label.txt file.')
+        exit(1)
+
+    if len(sys.argv) == 2:
+        if not os.path.isdir('widerface'):
+            os.mkdir('widerface')
+        if not os.path.isdir('widerface/val'):
+            os.mkdir('widerface/val')
+
+        save_path = 'widerface/val'
+    else:
+        save_path = sys.argv[2]
+
+    datas = wider2face(root_path, phase='val')
+    for idx, data in enumerate(datas.keys()):
+        pict_name = os.path.basename(data)
+        out_img = f'{save_path}/{idx}.jpg'
+        out_txt = f'{save_path}/{idx}.txt'
+        shutil.copyfile(data, out_img)
+        labels = datas[data]
+        f = open(out_txt, 'w')
+        for label in labels:
+            f.write(label + '\n')
+        f.close()

data/val2yolo_for_test.py

@@ -0,0 +1,65 @@
+import os
+import cv2
+import numpy as np
+import shutil
+from tqdm import tqdm
+
+root = '/ssd_1t/derron/WiderFace'
+
+
+def xywh2xxyy(box):
+    x1 = box[0]
+    y1 = box[1]
+    x2 = box[0] + box[2]
+    y2 = box[1] + box[3]
+    return (x1, x2, y1, y2)
+
+
+def convert(size, box):
+    dw = 1. / (size[0])
+    dh = 1. / (size[1])
+    x = (box[0] + box[1]) / 2.0 - 1
+    y = (box[2] + box[3]) / 2.0 - 1
+    w = box[1] - box[0]
+    h = box[3] - box[2]
+    x = x * dw
+    w = w * dw
+    y = y * dh
+    h = h * dh
+    return (x, y, w, h)
+
+
+def wider2face(phase='val', ignore_small=0):
+    data = {}
+    with open('{}/{}/label.txt'.format(root, phase), 'r') as f:
+        lines = f.readlines()
+        for line in tqdm(lines):
+            line = line.strip()
+            if '#' in line:
+                path = '{}/{}/images/{}'.format(root, phase, os.path.basename(line))
+                img = cv2.imread(path)
+                height, width, _ = img.shape
+                data[path] = list()
+            else:
+                box = np.array(line.split()[0:4], dtype=np.float32)  # (x1,y1,w,h)
+                if box[2] < ignore_small or box[3] < ignore_small:
+                    continue
+                box = convert((width, height), xywh2xxyy(box))
+                label = '0 {} {} {} {} -1 -1 -1 -1 -1 -1 -1 -1 -1 -1'.format(round(box[0], 4), round(box[1], 4),
+                                                                             round(box[2], 4), round(box[3], 4))
+                data[path].append(label)
+    return data
+
+
+if __name__ == '__main__':
+    datas = wider2face('val')
+    for idx, data in enumerate(datas.keys()):
+        pict_name = os.path.basename(data)
+        out_img = 'widerface/val/images/{}'.format(pict_name)
+        out_txt = 'widerface/val/labels/{}.txt'.format(os.path.splitext(pict_name)[0])
+        shutil.copyfile(data, out_img)
+        labels = datas[data]
+        f = open(out_txt, 'w')
+        for label in labels:
+            f.write(label + '\n')
+        f.close()

data/voc.yaml

@@ -0,0 +1,21 @@
+# PASCAL VOC dataset http://host.robots.ox.ac.uk/pascal/VOC/
+# Train command: python train.py --data voc.yaml
+# Default dataset location is next to /yolov5:
+#   /parent_folder
+#     /VOC
+#     /yolov5
+
+
+# download command/URL (optional)
+download: bash data/scripts/get_voc.sh
+
+# train and val data as 1) directory: path/images/, 2) file: path/images.txt, or 3) list: [path1/images/, path2/images/]
+train: ../VOC/images/train/  # 16551 images
+val: ../VOC/images/val/  # 4952 images
+
+# number of classes
+nc: 20
+
+# class names
+names: [ 'aeroplane', 'bicycle', 'bird', 'boat', 'bottle', 'bus', 'car', 'cat', 'chair', 'cow', 'diningtable', 'dog',
+         'horse', 'motorbike', 'person', 'pottedplant', 'sheep', 'sofa', 'train', 'tvmonitor' ]

data/widerface.yaml

@@ -0,0 +1,19 @@
+# PASCAL VOC dataset http://host.robots.ox.ac.uk/pascal/VOC/
+# Train command: python train.py --data voc.yaml
+# Default dataset location is next to /yolov5:
+#   /parent_folder
+#     /VOC
+#     /yolov5
+
+
+# download command/URL (optional)
+download: bash data/scripts/get_voc.sh
+
+# train and val data as 1) directory: path/images/, 2) file: path/images.txt, or 3) list: [path1/images/, path2/images/]
+train: /mnt/Gpan/Mydata/pytorchPorject/yolov5-face/ccpd/train_detect  
+val: /mnt/Gpan/Mydata/pytorchPorject/yolov5-face/ccpd/val_detect 
+# number of classes
+nc: 2
+
+# class names
+names: [ 'single','double']

demo.sh

@@ -0,0 +1 @@
+python detect_plate.py --detect_model runs/train/exp22/weights/last.pt --rec_model /mnt/Gpan/Mydata/pytorchPorject/CRNN/newCrnn/crnn_plate_recognition/output/360CC/crnn/2022-12-02-22-29/checkpoints/checkpoint_71_acc_0.9524.pth --image_path mytest --img_size 384

detect_demo.py

@@ -0,0 +1,223 @@
+# -*- coding: UTF-8 -*-
+import argparse
+import time
+import os
+import cv2
+import torch
+from numpy import random
+import copy
+import numpy as np
+from models.experimental import attempt_load
+from utils.datasets import letterbox
+from utils.general import check_img_size, non_max_suppression_face, scale_coords
+
+from utils.torch_utils import  time_synchronized
+from utils.cv_puttext import cv2ImgAddText
+from plate_recognition.plate_rec import get_plate_result,allFilePath,cv_imread
+
+from plate_recognition.double_plate_split_merge import get_split_merge
+
+clors = [(255,0,0),(0,255,0),(0,0,255),(255,255,0),(0,255,255)]
+
+def load_model(weights, device):
+    model = attempt_load(weights, map_location=device)  # load FP32 model
+    return model
+
+
+def scale_coords_landmarks(img1_shape, coords, img0_shape, ratio_pad=None):
+    # Rescale coords (xyxy) from img1_shape to img0_shape
+    if ratio_pad is None:  # calculate from img0_shape
+        gain = min(img1_shape[0] / img0_shape[0], img1_shape[1] / img0_shape[1])  # gain  = old / new
+        pad = (img1_shape[1] - img0_shape[1] * gain) / 2, (img1_shape[0] - img0_shape[0] * gain) / 2  # wh padding
+    else:
+        gain = ratio_pad[0][0]
+        pad = ratio_pad[1]
+
+    coords[:, [0, 2, 4, 6]] -= pad[0]  # x padding
+    coords[:, [1, 3, 5, 7]] -= pad[1]  # y padding
+    coords[:, :10] /= gain
+    #clip_coords(coords, img0_shape)
+    coords[:, 0].clamp_(0, img0_shape[1])  # x1
+    coords[:, 1].clamp_(0, img0_shape[0])  # y1
+    coords[:, 2].clamp_(0, img0_shape[1])  # x2
+    coords[:, 3].clamp_(0, img0_shape[0])  # y2
+    coords[:, 4].clamp_(0, img0_shape[1])  # x3
+    coords[:, 5].clamp_(0, img0_shape[0])  # y3
+    coords[:, 6].clamp_(0, img0_shape[1])  # x4
+    coords[:, 7].clamp_(0, img0_shape[0])  # y4
+    # coords[:, 8].clamp_(0, img0_shape[1])  # x5
+    # coords[:, 9].clamp_(0, img0_shape[0])  # y5
+    return coords
+
+
+
+
+def get_plate_rec_landmark(img, xyxy, conf, landmarks, class_num,device):
+    h,w,c = img.shape
+    result_dict={}
+    tl = 1 or round(0.002 * (h + w) / 2) + 1  # line/font thickness
+
+    x1 = int(xyxy[0])
+    y1 = int(xyxy[1])
+    x2 = int(xyxy[2])
+    y2 = int(xyxy[3])
+    landmarks_np=np.zeros((4,2))
+    rect=[x1,y1,x2,y2]
+    for i in range(4):
+        point_x = int(landmarks[2 * i])
+        point_y = int(landmarks[2 * i + 1])
+        landmarks_np[i]=np.array([point_x,point_y])
+
+    class_label= int(class_num)  #车牌的的类型0代表单牌，1代表双层车牌
+    result_dict['rect']=rect
+    result_dict['landmarks']=landmarks_np.tolist()
+    result_dict['class']=class_label
+    return result_dict
+
+
+
+def detect_plate(model, orgimg, device,img_size):
+    # Load model
+    # img_size = opt_img_size
+    conf_thres = 0.3
+    iou_thres = 0.5
+    dict_list=[]
+    # orgimg = cv2.imread(image_path)  # BGR
+    img0 = copy.deepcopy(orgimg)
+    assert orgimg is not None, 'Image Not Found ' 
+    h0, w0 = orgimg.shape[:2]  # orig hw
+    r = img_size / max(h0, w0)  # resize image to img_size
+    if r != 1:  # always resize down, only resize up if training with augmentation
+        interp = cv2.INTER_AREA if r < 1  else cv2.INTER_LINEAR
+        img0 = cv2.resize(img0, (int(w0 * r), int(h0 * r)), interpolation=interp)
+
+    imgsz = check_img_size(img_size, s=model.stride.max())  # check img_size
+
+    img = letterbox(img0, new_shape=imgsz)[0]
+    # img =process_data(img0)
+    # Convert
+    img = img[:, :, ::-1].transpose(2, 0, 1).copy()  # BGR to RGB, to 3x416x416
+
+    # Run inference
+    t0 = time.time()
+
+    img = torch.from_numpy(img).to(device)
+    img = img.float()  # uint8 to fp16/32
+    img /= 255.0  # 0 - 255 to 0.0 - 1.0
+    if img.ndimension() == 3:
+        img = img.unsqueeze(0)
+
+    # Inference
+    t1 = time_synchronized()
+    pred = model(img)[0]
+    t2=time_synchronized()
+    # print(f"infer time is {(t2-t1)*1000} ms")
+
+    # Apply NMS
+    pred = non_max_suppression_face(pred, conf_thres, iou_thres)
+
+    # print('img.shape: ', img.shape)
+    # print('orgimg.shape: ', orgimg.shape)
+
+    # Process detections
+    for i, det in enumerate(pred):  # detections per image
+        if len(det):
+            # Rescale boxes from img_size to im0 size
+            det[:, :4] = scale_coords(img.shape[2:], det[:, :4], orgimg.shape).round()
+
+            # Print results
+            for c in det[:, -1].unique():
+                n = (det[:, -1] == c).sum()  # detections per class
+
+            det[:, 5:13] = scale_coords_landmarks(img.shape[2:], det[:, 5:13], orgimg.shape).round()
+
+            for j in range(det.size()[0]):
+                xyxy = det[j, :4].view(-1).tolist()
+                conf = det[j, 4].cpu().numpy()
+                landmarks = det[j, 5:13].view(-1).tolist()
+                class_num = det[j, 13].cpu().numpy()
+                result_dict = get_plate_rec_landmark(orgimg, xyxy, conf, landmarks, class_num,device)
+                dict_list.append(result_dict)
+    return dict_list
+    # cv2.imwrite('result.jpg', orgimg)
+
+
+
+def draw_result(orgimg,dict_list):
+    result_str =""
+    for result in dict_list:
+        rect_area = result['rect']
+        
+        x,y,w,h = rect_area[0],rect_area[1],rect_area[2]-rect_area[0],rect_area[3]-rect_area[1]
+        padding_w = 0.05*w
+        padding_h = 0.11*h
+        rect_area[0]=max(0,int(x-padding_w))
+        rect_area[1]=max(0,int(y-padding_h))
+        rect_area[2]=min(orgimg.shape[1],int(rect_area[2]+padding_w))
+        rect_area[3]=min(orgimg.shape[0],int(rect_area[3]+padding_h))
+
+        
+        landmarks=result['landmarks']
+        label=result['class']
+        # result_str+=result+" "
+        for i in range(4):  #关键点
+            cv2.circle(orgimg, (int(landmarks[i][0]), int(landmarks[i][1])), 5, clors[i], -1)
+        cv2.rectangle(orgimg,(rect_area[0],rect_area[1]),(rect_area[2],rect_area[3]),clors[label],2) #画框
+        cv2.putText(img,str(label),(rect_area[0],rect_area[1]),cv2.FONT_HERSHEY_SIMPLEX,0.5,clors[label],2)
+    #     orgimg=cv2ImgAddText(orgimg,label,rect_area[0]-height_area,rect_area[1]-height_area-10,(0,255,0),height_area)
+    # print(result_str)
+    return orgimg
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--detect_model', nargs='+', type=str, default='weights/plate_detect.pt', help='model.pt path(s)')  #检测模型
+    parser.add_argument('--image_path', type=str, default='imgs', help='source') 
+    parser.add_argument('--img_size', type=int, default=640, help='inference size (pixels)')
+    parser.add_argument('--output', type=str, default='result1', help='source') 
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+    # device =torch.device("cpu")
+    opt = parser.parse_args()
+    print(opt)
+    save_path = opt.output
+    count=0
+    if not os.path.exists(save_path):
+        os.mkdir(save_path)
+
+    detect_model = load_model(opt.detect_model, device)  #初始化检测模型
+    time_all = 0
+    time_begin=time.time()
+    if not os.path.isfile(opt.image_path):            #目录
+        file_list=[]
+        allFilePath(opt.image_path,file_list)
+        for img_path in file_list:
+            
+            print(count,img_path)
+            time_b = time.time()
+            img =cv_imread(img_path)
+            
+            if img is None:
+                continue
+            if img.shape[-1]==4:
+                img=cv2.cvtColor(img,cv2.COLOR_BGRA2BGR)
+            # detect_one(model,img_path,device)
+            dict_list=detect_plate(detect_model, img, device,opt.img_size)
+            ori_img=draw_result(img,dict_list)
+            img_name = os.path.basename(img_path)
+            save_img_path = os.path.join(save_path,img_name)
+            time_e=time.time()
+            time_gap = time_e-time_b
+            if count:
+                time_all+=time_gap
+            cv2.imwrite(save_img_path,ori_img)
+            count+=1
+    else:                                          #单个图片
+            print(count,opt.image_path,end=" ")
+            img =cv_imread(opt.image_path)
+            if img.shape[-1]==4:
+                img=cv2.cvtColor(img,cv2.COLOR_BGRA2BGR)
+            # detect_one(model,img_path,device)
+            dict_list=detect_plate(detect_model, img, device,opt.img_size)
+            ori_img=draw_result(img,dict_list)
+            img_name = os.path.basename(opt.image_path)
+            save_img_path = os.path.join(save_path,img_name)
+            cv2.imwrite(save_img_path,ori_img)  
+    print(f"sumTime time is {time.time()-time_begin} s, average pic time is {time_all/(len(file_list)-1)}")

export.py

@@ -0,0 +1,161 @@
+"""Exports a YOLOv5 *.pt model to ONNX and TorchScript formats
+
+Usage:
+    $ export PYTHONPATH="$PWD" && python models/export.py --weights ./weights/yolov5s.pt --img 640 --batch 1
+"""
+
+import argparse
+import sys
+import time
+
+sys.path.append('./')  # to run '$ python *.py' files in subdirectories
+
+import torch
+import torch.nn as nn
+
+import models
+from models.experimental import attempt_load
+from utils.activations import Hardswish, SiLU
+from utils.general import set_logging, check_img_size
+import onnx
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--weights', type=str, default='./yolov5s.pt', help='weights path')  # from yolov5/models/
+    parser.add_argument('--img_size', nargs='+', type=int, default=[640, 640], help='image size')  # height, width
+    parser.add_argument('--batch_size', type=int, default=1, help='batch size')
+    parser.add_argument('--dynamic', action='store_true', default=False, help='enable dynamic axis in onnx model')
+    parser.add_argument('--onnx2pb', action='store_true', default=False, help='export onnx to pb')
+    parser.add_argument('--onnx_infer', action='store_true', default=True, help='onnx infer test')
+    #=======================TensorRT=================================
+    parser.add_argument('--onnx2trt', action='store_true', default=False, help='export onnx to tensorrt')
+    parser.add_argument('--fp16_trt', action='store_true', default=False, help='fp16 infer')
+    #================================================================
+    opt = parser.parse_args()
+    opt.img_size *= 2 if len(opt.img_size) == 1 else 1  # expand
+    print(opt)
+    set_logging()
+    t = time.time()
+
+    # Load PyTorch model
+    model = attempt_load(opt.weights, map_location=torch.device('cpu'))  # load FP32 model
+    delattr(model.model[-1], 'anchor_grid')
+    model.model[-1].anchor_grid=[torch.zeros(1)] * 3 # nl=3 number of detection layers
+    model.model[-1].export_cat = True
+    model.eval()
+    labels = model.names
+
+    # Checks
+    gs = int(max(model.stride))  # grid size (max stride)
+    opt.img_size = [check_img_size(x, gs) for x in opt.img_size]  # verify img_size are gs-multiples
+
+    # Input
+    img = torch.zeros(opt.batch_size, 3, *opt.img_size)  # image size(1,3,320,192) iDetection
+
+    # Update model
+    for k, m in model.named_modules():
+        m._non_persistent_buffers_set = set()  # pytorch 1.6.0 compatibility
+        if isinstance(m, models.common.Conv):  # assign export-friendly activations
+            if isinstance(m.act, nn.Hardswish):
+                m.act = Hardswish()
+            elif isinstance(m.act, nn.SiLU):
+                m.act = SiLU()
+        # elif isinstance(m, models.yolo.Detect):
+        #     m.forward = m.forward_export  # assign forward (optional)
+        if isinstance(m, models.common.ShuffleV2Block):#shufflenet block nn.SiLU
+            for i in range(len(m.branch1)):
+                if isinstance(m.branch1[i], nn.SiLU):
+                    m.branch1[i] = SiLU()
+            for i in range(len(m.branch2)):
+                if isinstance(m.branch2[i], nn.SiLU):
+                    m.branch2[i] = SiLU()
+        if isinstance(m, models.common.BlazeBlock):#shufflenet block nn.SiLU
+            if isinstance(m.relu, nn.SiLU):
+                m.relu = SiLU()
+        if isinstance(m, models.common.DoubleBlazeBlock):#shufflenet block nn.SiLU
+            if isinstance(m.relu, nn.SiLU):
+                m.relu = SiLU()
+            for i in range(len(m.branch1)):
+                if isinstance(m.branch1[i], nn.SiLU):
+                    m.branch1[i] = SiLU()
+            # for i in range(len(m.branch2)):
+            #     if isinstance(m.branch2[i], nn.SiLU):
+            #         m.branch2[i] = SiLU()
+    y = model(img)  # dry run
+
+    # ONNX export
+    print('\nStarting ONNX export with onnx %s...' % onnx.__version__)
+    f = opt.weights.replace('.pt', '.onnx')  # filename
+    model.fuse()  # only for ONNX
+    input_names=['input']
+    output_names=['output']
+    #tensorrt 7
+    # grid = model.model[-1].anchor_grid
+    # model.model[-1].anchor_grid = [a[..., :1, :1, :] for a in grid]
+    #tensorrt 7
+
+    torch.onnx.export(model, img, f, verbose=False, opset_version=12, 
+        input_names=input_names,
+        output_names=output_names,
+        dynamic_axes = {'input': {0: 'batch'},
+                        'output': {0: 'batch'}
+                        } if opt.dynamic else None)
+                        
+    # model.model[-1].anchor_grid = grid
+
+    # Checks
+    onnx_model = onnx.load(f)  # load onnx model
+    onnx.checker.check_model(onnx_model)  # check onnx model
+    print('ONNX export success, saved as %s' % f)
+    # Finish
+    print('\nExport complete (%.2fs). Visualize with https://github.com/lutzroeder/netron.' % (time.time() - t))
+
+
+    # onnx infer
+    if opt.onnx_infer:
+        import onnxruntime
+        import numpy as np
+        providers =  ['CPUExecutionProvider']
+        session = onnxruntime.InferenceSession(f, providers=providers)
+        im = img.cpu().numpy().astype(np.float32) # torch to numpy
+        y_onnx = session.run([session.get_outputs()[0].name], {session.get_inputs()[0].name: im})[0]
+        print("pred's shape is ",y_onnx.shape)
+        print("max(|torch_pred - onnx_pred|） =",abs(y.cpu().numpy()-y_onnx).max())
+
+
+    # TensorRT export
+    if opt.onnx2trt:
+        from torch2trt.trt_model import ONNX_to_TRT
+        print('\nStarting TensorRT...')
+        ONNX_to_TRT(onnx_model_path=f,trt_engine_path=f.replace('.onnx', '.trt'),fp16_mode=opt.fp16_trt)
+
+    # PB export
+    if opt.onnx2pb:
+        print('download the newest onnx_tf by https://github.com/onnx/onnx-tensorflow/tree/master/onnx_tf')
+        from onnx_tf.backend import prepare
+        import tensorflow as tf
+
+        outpb = f.replace('.onnx', '.pb')  # filename
+        # strict=True maybe leads to KeyError: 'pyfunc_0', check: https://github.com/onnx/onnx-tensorflow/issues/167
+        tf_rep = prepare(onnx_model, strict=False)  # prepare tf representation
+        tf_rep.export_graph(outpb)  # export the model
+
+        out_onnx = tf_rep.run(img) # onnx output
+
+        # check pb
+        with tf.Graph().as_default():
+            graph_def = tf.GraphDef()
+            with open(outpb, "rb") as f:
+                graph_def.ParseFromString(f.read())
+                tf.import_graph_def(graph_def, name="")
+            with tf.Session() as sess:
+                init = tf.global_variables_initializer()
+                input_x = sess.graph.get_tensor_by_name(input_names[0]+':0')  # input
+                outputs = []
+                for i in output_names:
+                    outputs.append(sess.graph.get_tensor_by_name(i+':0'))
+                out_pb = sess.run(outputs, feed_dict={input_x: img})
+
+        print(f'out_pytorch {y}')
+        print(f'out_onnx {out_onnx}')
+        print(f'out_pb {out_pb}')

hubconf.py

@@ -0,0 +1,141 @@
+"""File for accessing YOLOv5 via PyTorch Hub https://pytorch.org/hub/
+
+Usage:
+    import torch
+    model = torch.hub.load('ultralytics/yolov5', 'yolov5s', pretrained=True, channels=3, classes=80)
+"""
+
+from pathlib import Path
+
+import torch
+
+from models.yolo import Model
+from utils.general import set_logging
+from utils.google_utils import attempt_download
+
+dependencies = ['torch', 'yaml']
+set_logging()
+
+
+def create(name, pretrained, channels, classes, autoshape):
+    """Creates a specified YOLOv5 model
+
+    Arguments:
+        name (str): name of model, i.e. 'yolov5s'
+        pretrained (bool): load pretrained weights into the model
+        channels (int): number of input channels
+        classes (int): number of model classes
+
+    Returns:
+        pytorch model
+    """
+    config = Path(__file__).parent / 'models' / f'{name}.yaml'  # model.yaml path
+    try:
+        model = Model(config, channels, classes)
+        if pretrained:
+            fname = f'{name}.pt'  # checkpoint filename
+            attempt_download(fname)  # download if not found locally
+            ckpt = torch.load(fname, map_location=torch.device('cpu'))  # load
+            state_dict = ckpt['model'].float().state_dict()  # to FP32
+            state_dict = {k: v for k, v in state_dict.items() if model.state_dict()[k].shape == v.shape}  # filter
+            model.load_state_dict(state_dict, strict=False)  # load
+            if len(ckpt['model'].names) == classes:
+                model.names = ckpt['model'].names  # set class names attribute
+            if autoshape:
+                model = model.autoshape()  # for file/URI/PIL/cv2/np inputs and NMS
+        return model
+
+    except Exception as e:
+        help_url = 'https://github.com/ultralytics/yolov5/issues/36'
+        s = 'Cache maybe be out of date, try force_reload=True. See %s for help.' % help_url
+        raise Exception(s) from e
+
+
+def yolov5s(pretrained=False, channels=3, classes=80, autoshape=True):
+    """YOLOv5-small model from https://github.com/ultralytics/yolov5
+
+    Arguments:
+        pretrained (bool): load pretrained weights into the model, default=False
+        channels (int): number of input channels, default=3
+        classes (int): number of model classes, default=80
+
+    Returns:
+        pytorch model
+    """
+    return create('yolov5s', pretrained, channels, classes, autoshape)
+
+
+def yolov5m(pretrained=False, channels=3, classes=80, autoshape=True):
+    """YOLOv5-medium model from https://github.com/ultralytics/yolov5
+
+    Arguments:
+        pretrained (bool): load pretrained weights into the model, default=False
+        channels (int): number of input channels, default=3
+        classes (int): number of model classes, default=80
+
+    Returns:
+        pytorch model
+    """
+    return create('yolov5m', pretrained, channels, classes, autoshape)
+
+
+def yolov5l(pretrained=False, channels=3, classes=80, autoshape=True):
+    """YOLOv5-large model from https://github.com/ultralytics/yolov5
+
+    Arguments:
+        pretrained (bool): load pretrained weights into the model, default=False
+        channels (int): number of input channels, default=3
+        classes (int): number of model classes, default=80
+
+    Returns:
+        pytorch model
+    """
+    return create('yolov5l', pretrained, channels, classes, autoshape)
+
+
+def yolov5x(pretrained=False, channels=3, classes=80, autoshape=True):
+    """YOLOv5-xlarge model from https://github.com/ultralytics/yolov5
+
+    Arguments:
+        pretrained (bool): load pretrained weights into the model, default=False
+        channels (int): number of input channels, default=3
+        classes (int): number of model classes, default=80
+
+    Returns:
+        pytorch model
+    """
+    return create('yolov5x', pretrained, channels, classes, autoshape)
+
+
+def custom(path_or_model='path/to/model.pt', autoshape=True):
+    """YOLOv5-custom model from https://github.com/ultralytics/yolov5
+
+    Arguments (3 options):
+        path_or_model (str): 'path/to/model.pt'
+        path_or_model (dict): torch.load('path/to/model.pt')
+        path_or_model (nn.Module): torch.load('path/to/model.pt')['model']
+
+    Returns:
+        pytorch model
+    """
+    model = torch.load(path_or_model) if isinstance(path_or_model, str) else path_or_model  # load checkpoint
+    if isinstance(model, dict):
+        model = model['model']  # load model
+
+    hub_model = Model(model.yaml).to(next(model.parameters()).device)  # create
+    hub_model.load_state_dict(model.float().state_dict())  # load state_dict
+    hub_model.names = model.names  # class names
+    return hub_model.autoshape() if autoshape else hub_model
+
+
+if __name__ == '__main__':
+    model = create(name='yolov5s', pretrained=True, channels=3, classes=80, autoshape=True)  # pretrained example
+    # model = custom(path_or_model='path/to/model.pt')  # custom example
+
+    # Verify inference
+    from PIL import Image
+
+    imgs = [Image.open(x) for x in Path('data/images').glob('*.jpg')]
+    results = model(imgs)
+    results.show()
+    results.print()

models/blazeface.yaml

@@ -0,0 +1,33 @@
+# parameters
+nc: 1  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+
+# anchors
+anchors:
+  - [5,6,  10,13,  21,26]  # P3/8
+  - [55,72,  225,304,  438,553]  # P4/16
+
+# YOLOv5 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [24, 3, 2]], # 0-P1/2
+   [-1, 2, BlazeBlock, [24]], # 1
+   [-1, 1, BlazeBlock, [48, None, 2]], # 2-P2/4
+   [-1, 2, BlazeBlock, [48]], # 3
+   [-1, 1, DoubleBlazeBlock, [96, 24, 2]], # 4-P3/8
+   [-1, 2, DoubleBlazeBlock, [96, 24]], # 5
+   [-1, 1, DoubleBlazeBlock, [96, 24, 2]], # 6-P4/16
+   [-1, 2, DoubleBlazeBlock, [96, 24]], # 7
+  ]
+
+
+# YOLOv5 head
+head:
+  [[-1, 1, Conv, [64, 1, 1]],  # 8 (P4/32-large)
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 5], 1, Concat, [1]],  # cat backbone P3
+   [-1, 1, Conv, [64, 1, 1]],  # 11 (P3/8-medium)
+
+   [[11, 8], 1, Detect, [nc, anchors]],  # Detect(P3, P4)
+  ]

models/blazeface_fpn.yaml

@@ -0,0 +1,38 @@
+# parameters
+nc: 1  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+
+# anchors
+anchors:
+  - [5,6,  10,13,  21,26]  # P3/8
+  - [55,72,  225,304,  438,553]  # P4/16
+
+# YOLOv5 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, Conv, [24, 3, 2]], # 0-P1/2
+   [-1, 2, BlazeBlock, [24]], # 1
+   [-1, 1, BlazeBlock, [48, None, 2]], # 2-P2/4
+   [-1, 2, BlazeBlock, [48]], # 3
+   [-1, 1, DoubleBlazeBlock, [96, 24, 2]], # 4-P3/8
+   [-1, 2, DoubleBlazeBlock, [96, 24]], # 5
+   [-1, 1, DoubleBlazeBlock, [96, 24, 2]], # 6-P4/16
+   [-1, 2, DoubleBlazeBlock, [96, 24]], # 7
+  ]
+
+
+# YOLOv5 head
+head:
+  [[-1, 1, Conv, [48, 1, 1]],  # 8
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 5], 1, Concat, [1]],  # cat backbone P3
+   [-1, 1, Conv, [48, 1, 1]],  # 11 (P3/8-medium)
+
+   [-1, 1, nn.MaxPool2d, [3, 2, 1]],  # 12
+   [[-1, 7], 1, Concat, [1]],  # cat backbone P3
+   [-1, 1, Conv, [48, 1, 1]],  # 14 (P4/16-large)
+
+   [[11, 14], 1, Detect, [nc, anchors]],  # Detect(P3, P4)
+  ]
+

models/common.py

@@ -0,0 +1,456 @@
+# This file contains modules common to various models
+
+import math
+
+import numpy as np
+import requests
+import torch
+import torch.nn as nn
+from PIL import Image, ImageDraw
+
+from utils.datasets import letterbox
+from utils.general import non_max_suppression, make_divisible, scale_coords, xyxy2xywh
+from utils.plots import color_list
+
+def autopad(k, p=None):  # kernel, padding
+    # Pad to 'same'
+    if p is None:
+        p = k // 2 if isinstance(k, int) else [x // 2 for x in k]  # auto-pad
+    return p
+
+def channel_shuffle(x, groups):
+    batchsize, num_channels, height, width = x.data.size()
+    channels_per_group = num_channels // groups
+
+    # reshape
+    x = x.view(batchsize, groups, channels_per_group, height, width)
+    x = torch.transpose(x, 1, 2).contiguous()
+
+    # flatten
+    x = x.view(batchsize, -1, height, width)
+    return x
+
+def DWConv(c1, c2, k=1, s=1, act=True):
+    # Depthwise convolution
+    return Conv(c1, c2, k, s, g=math.gcd(c1, c2), act=act)
+
+class Conv(nn.Module):
+    # Standard convolution
+    def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True):  # ch_in, ch_out, kernel, stride, padding, groups
+        super(Conv, self).__init__()
+        self.conv = nn.Conv2d(c1, c2, k, s, autopad(k, p), groups=g, bias=False)
+        self.bn = nn.BatchNorm2d(c2)
+        self.act = nn.SiLU() if act is True else (act if isinstance(act, nn.Module) else nn.Identity())
+        #self.act = self.act = nn.LeakyReLU(0.1, inplace=True) if act is True else (act if isinstance(act, nn.Module) else nn.Identity())
+
+    def forward(self, x):
+        return self.act(self.bn(self.conv(x)))
+
+    def fuseforward(self, x):
+        return self.act(self.conv(x))
+
+class StemBlock(nn.Module):
+    def __init__(self, c1, c2, k=3, s=2, p=None, g=1, act=True):
+        super(StemBlock, self).__init__()
+        self.stem_1 = Conv(c1, c2, k, s, p, g, act)
+        self.stem_2a = Conv(c2, c2 // 2, 1, 1, 0)
+        self.stem_2b = Conv(c2 // 2, c2, 3, 2, 1)
+        self.stem_2p = nn.MaxPool2d(kernel_size=2,stride=2,ceil_mode=True)
+        self.stem_3 = Conv(c2 * 2, c2, 1, 1, 0)
+
+    def forward(self, x):
+        stem_1_out  = self.stem_1(x)
+        stem_2a_out = self.stem_2a(stem_1_out)
+        stem_2b_out = self.stem_2b(stem_2a_out)
+        stem_2p_out = self.stem_2p(stem_1_out)
+        out = self.stem_3(torch.cat((stem_2b_out,stem_2p_out),1))
+        return out
+
+class Bottleneck(nn.Module):
+    # Standard bottleneck
+    def __init__(self, c1, c2, shortcut=True, g=1, e=0.5):  # ch_in, ch_out, shortcut, groups, expansion
+        super(Bottleneck, self).__init__()
+        c_ = int(c2 * e)  # hidden channels
+        self.cv1 = Conv(c1, c_, 1, 1)
+        self.cv2 = Conv(c_, c2, 3, 1, g=g)
+        self.add = shortcut and c1 == c2
+
+    def forward(self, x):
+        return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x))
+
+class BottleneckCSP(nn.Module):
+    # CSP Bottleneck https://github.com/WongKinYiu/CrossStagePartialNetworks
+    def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):  # ch_in, ch_out, number, shortcut, groups, expansion
+        super(BottleneckCSP, self).__init__()
+        c_ = int(c2 * e)  # hidden channels
+        self.cv1 = Conv(c1, c_, 1, 1)
+        self.cv2 = nn.Conv2d(c1, c_, 1, 1, bias=False)
+        self.cv3 = nn.Conv2d(c_, c_, 1, 1, bias=False)
+        self.cv4 = Conv(2 * c_, c2, 1, 1)
+        self.bn = nn.BatchNorm2d(2 * c_)  # applied to cat(cv2, cv3)
+        self.act = nn.LeakyReLU(0.1, inplace=True)
+        self.m = nn.Sequential(*[Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n)])
+
+    def forward(self, x):
+        y1 = self.cv3(self.m(self.cv1(x)))
+        y2 = self.cv2(x)
+        return self.cv4(self.act(self.bn(torch.cat((y1, y2), dim=1))))
+
+
+class C3(nn.Module):
+    # CSP Bottleneck with 3 convolutions
+    def __init__(self, c1, c2, n=1, shortcut=True, g=1, e=0.5):  # ch_in, ch_out, number, shortcut, groups, expansion
+        super(C3, self).__init__()
+        c_ = int(c2 * e)  # hidden channels
+        self.cv1 = Conv(c1, c_, 1, 1)
+        self.cv2 = Conv(c1, c_, 1, 1)
+        self.cv3 = Conv(2 * c_, c2, 1)  # act=FReLU(c2)
+        self.m = nn.Sequential(*[Bottleneck(c_, c_, shortcut, g, e=1.0) for _ in range(n)])
+
+    def forward(self, x):
+        return self.cv3(torch.cat((self.m(self.cv1(x)), self.cv2(x)), dim=1))
+
+class ShuffleV2Block(nn.Module):
+    def __init__(self, inp, oup, stride):
+        super(ShuffleV2Block, self).__init__()
+
+        if not (1 <= stride <= 3):
+            raise ValueError('illegal stride value')
+        self.stride = stride
+
+        branch_features = oup // 2
+        assert (self.stride != 1) or (inp == branch_features << 1)
+
+        if self.stride > 1:
+            self.branch1 = nn.Sequential(
+                self.depthwise_conv(inp, inp, kernel_size=3, stride=self.stride, padding=1),
+                nn.BatchNorm2d(inp),
+                nn.Conv2d(inp, branch_features, kernel_size=1, stride=1, padding=0, bias=False),
+                nn.BatchNorm2d(branch_features),
+                nn.SiLU(),
+            )
+        else:
+            self.branch1 = nn.Sequential()
+
+        self.branch2 = nn.Sequential(
+            nn.Conv2d(inp if (self.stride > 1) else branch_features, branch_features, kernel_size=1, stride=1, padding=0, bias=False),
+            nn.BatchNorm2d(branch_features),
+            nn.SiLU(),
+            self.depthwise_conv(branch_features, branch_features, kernel_size=3, stride=self.stride, padding=1),
+            nn.BatchNorm2d(branch_features),
+            nn.Conv2d(branch_features, branch_features, kernel_size=1, stride=1, padding=0, bias=False),
+            nn.BatchNorm2d(branch_features),
+            nn.SiLU(),
+        )
+
+    @staticmethod
+    def depthwise_conv(i, o, kernel_size, stride=1, padding=0, bias=False):
+        return nn.Conv2d(i, o, kernel_size, stride, padding, bias=bias, groups=i)
+
+    def forward(self, x):
+        if self.stride == 1:
+            x1, x2 = x.chunk(2, dim=1)
+            out = torch.cat((x1, self.branch2(x2)), dim=1)
+        else:
+            out = torch.cat((self.branch1(x), self.branch2(x)), dim=1)
+        out = channel_shuffle(out, 2)
+        return out
+    
+class BlazeBlock(nn.Module):
+    def __init__(self, in_channels,out_channels,mid_channels=None,stride=1):
+        super(BlazeBlock, self).__init__()
+        mid_channels = mid_channels or in_channels
+        assert stride in [1, 2]
+        if stride>1:
+            self.use_pool = True
+        else:
+            self.use_pool = False
+
+        self.branch1 = nn.Sequential(
+            nn.Conv2d(in_channels=in_channels,out_channels=mid_channels,kernel_size=5,stride=stride,padding=2,groups=in_channels),
+            nn.BatchNorm2d(mid_channels),
+            nn.Conv2d(in_channels=mid_channels,out_channels=out_channels,kernel_size=1,stride=1),
+            nn.BatchNorm2d(out_channels),
+        )
+
+        if self.use_pool:
+            self.shortcut = nn.Sequential(
+                nn.MaxPool2d(kernel_size=stride, stride=stride),
+                nn.Conv2d(in_channels=in_channels, out_channels=out_channels, kernel_size=1, stride=1),
+                nn.BatchNorm2d(out_channels),
+            )
+
+        self.relu = nn.SiLU(inplace=True)
+
+    def forward(self, x):
+        branch1 = self.branch1(x)
+        out = (branch1+self.shortcut(x)) if self.use_pool else (branch1+x)
+        return self.relu(out)    
+  
+class DoubleBlazeBlock(nn.Module):
+    def __init__(self,in_channels,out_channels,mid_channels=None,stride=1):
+        super(DoubleBlazeBlock, self).__init__()
+        mid_channels = mid_channels or in_channels
+        assert stride in [1, 2]
+        if stride > 1:
+            self.use_pool = True
+        else:
+            self.use_pool = False
+
+        self.branch1 = nn.Sequential(
+            nn.Conv2d(in_channels=in_channels, out_channels=in_channels, kernel_size=5, stride=stride,padding=2,groups=in_channels),
+            nn.BatchNorm2d(in_channels),
+            nn.Conv2d(in_channels=in_channels, out_channels=mid_channels, kernel_size=1, stride=1),
+            nn.BatchNorm2d(mid_channels),
+            nn.SiLU(inplace=True),
+            nn.Conv2d(in_channels=mid_channels, out_channels=mid_channels, kernel_size=5, stride=1,padding=2),
+            nn.BatchNorm2d(mid_channels),
+            nn.Conv2d(in_channels=mid_channels, out_channels=out_channels, kernel_size=1, stride=1),
+            nn.BatchNorm2d(out_channels),
+        )
+
+        if self.use_pool:
+            self.shortcut = nn.Sequential(
+                nn.MaxPool2d(kernel_size=stride, stride=stride),
+                nn.Conv2d(in_channels=in_channels, out_channels=out_channels, kernel_size=1, stride=1),
+                nn.BatchNorm2d(out_channels),
+            )
+
+        self.relu = nn.SiLU(inplace=True)
+
+    def forward(self, x):
+        branch1 = self.branch1(x)
+        out = (branch1 + self.shortcut(x)) if self.use_pool else (branch1 + x)
+        return self.relu(out)
+    
+    
+class SPP(nn.Module):
+    # Spatial pyramid pooling layer used in YOLOv3-SPP
+    def __init__(self, c1, c2, k=(5, 9, 13)):
+        super(SPP, self).__init__()
+        c_ = c1 // 2  # hidden channels
+        self.cv1 = Conv(c1, c_, 1, 1)
+        self.cv2 = Conv(c_ * (len(k) + 1), c2, 1, 1)
+        self.m = nn.ModuleList([nn.MaxPool2d(kernel_size=x, stride=1, padding=x // 2) for x in k])
+
+    def forward(self, x):
+        x = self.cv1(x)
+        return self.cv2(torch.cat([x] + [m(x) for m in self.m], 1))
+
+class SPPF(nn.Module):
+    # Spatial Pyramid Pooling - Fast (SPPF) layer for YOLOv5 by Glenn Jocher
+    def __init__(self, c1, c2, k=5):  # equivalent to SPP(k=(5, 9, 13))
+        super().__init__()
+        c_ = c1 // 2  # hidden channels
+        self.cv1 = Conv(c1, c_, 1, 1)
+        self.cv2 = Conv(c_ * 4, c2, 1, 1)
+        self.m = nn.MaxPool2d(kernel_size=k, stride=1, padding=k // 2)
+
+    def forward(self, x):
+        x = self.cv1(x)
+        with warnings.catch_warnings():
+            warnings.simplefilter('ignore')  # suppress torch 1.9.0 max_pool2d() warning
+            y1 = self.m(x)
+            y2 = self.m(y1)
+            return self.cv2(torch.cat((x, y1, y2, self.m(y2)), 1))
+
+
+class Focus(nn.Module):
+    # Focus wh information into c-space
+    def __init__(self, c1, c2, k=1, s=1, p=None, g=1, act=True):  # ch_in, ch_out, kernel, stride, padding, groups
+        super(Focus, self).__init__()
+        self.conv = Conv(c1 * 4, c2, k, s, p, g, act)
+        # self.contract = Contract(gain=2)
+
+    def forward(self, x):  # x(b,c,w,h) -> y(b,4c,w/2,h/2)
+        return self.conv(torch.cat([x[..., ::2, ::2], x[..., 1::2, ::2], x[..., ::2, 1::2], x[..., 1::2, 1::2]], 1))
+        # return self.conv(self.contract(x))
+
+
+class Contract(nn.Module):
+    # Contract width-height into channels, i.e. x(1,64,80,80) to x(1,256,40,40)
+    def __init__(self, gain=2):
+        super().__init__()
+        self.gain = gain
+
+    def forward(self, x):
+        N, C, H, W = x.size()  # assert (H / s == 0) and (W / s == 0), 'Indivisible gain'
+        s = self.gain
+        x = x.view(N, C, H // s, s, W // s, s)  # x(1,64,40,2,40,2)
+        x = x.permute(0, 3, 5, 1, 2, 4).contiguous()  # x(1,2,2,64,40,40)
+        return x.view(N, C * s * s, H // s, W // s)  # x(1,256,40,40)
+
+
+class Expand(nn.Module):
+    # Expand channels into width-height, i.e. x(1,64,80,80) to x(1,16,160,160)
+    def __init__(self, gain=2):
+        super().__init__()
+        self.gain = gain
+
+    def forward(self, x):
+        N, C, H, W = x.size()  # assert C / s ** 2 == 0, 'Indivisible gain'
+        s = self.gain
+        x = x.view(N, s, s, C // s ** 2, H, W)  # x(1,2,2,16,80,80)
+        x = x.permute(0, 3, 4, 1, 5, 2).contiguous()  # x(1,16,80,2,80,2)
+        return x.view(N, C // s ** 2, H * s, W * s)  # x(1,16,160,160)
+
+
+class Concat(nn.Module):
+    # Concatenate a list of tensors along dimension
+    def __init__(self, dimension=1):
+        super(Concat, self).__init__()
+        self.d = dimension
+
+    def forward(self, x):
+        return torch.cat(x, self.d)
+
+
+class NMS(nn.Module):
+    # Non-Maximum Suppression (NMS) module
+    conf = 0.25  # confidence threshold
+    iou = 0.45  # IoU threshold
+    classes = None  # (optional list) filter by class
+
+    def __init__(self):
+        super(NMS, self).__init__()
+
+    def forward(self, x):
+        return non_max_suppression(x[0], conf_thres=self.conf, iou_thres=self.iou, classes=self.classes)
+
+class autoShape(nn.Module):
+    # input-robust model wrapper for passing cv2/np/PIL/torch inputs. Includes preprocessing, inference and NMS
+    img_size = 640  # inference size (pixels)
+    conf = 0.25  # NMS confidence threshold
+    iou = 0.45  # NMS IoU threshold
+    classes = None  # (optional list) filter by class
+
+    def __init__(self, model):
+        super(autoShape, self).__init__()
+        self.model = model.eval()
+
+    def autoshape(self):
+        print('autoShape already enabled, skipping... ')  # model already converted to model.autoshape()
+        return self
+
+    def forward(self, imgs, size=640, augment=False, profile=False):
+        # Inference from various sources. For height=720, width=1280, RGB images example inputs are:
+        #   filename:   imgs = 'data/samples/zidane.jpg'
+        #   URI:             = 'https://github.com/ultralytics/yolov5/releases/download/v1.0/zidane.jpg'
+        #   OpenCV:          = cv2.imread('image.jpg')[:,:,::-1]  # HWC BGR to RGB x(720,1280,3)
+        #   PIL:             = Image.open('image.jpg')  # HWC x(720,1280,3)
+        #   numpy:           = np.zeros((720,1280,3))  # HWC
+        #   torch:           = torch.zeros(16,3,720,1280)  # BCHW
+        #   multiple:        = [Image.open('image1.jpg'), Image.open('image2.jpg'), ...]  # list of images
+
+        p = next(self.model.parameters())  # for device and type
+        if isinstance(imgs, torch.Tensor):  # torch
+            return self.model(imgs.to(p.device).type_as(p), augment, profile)  # inference
+
+        # Pre-process
+        n, imgs = (len(imgs), imgs) if isinstance(imgs, list) else (1, [imgs])  # number of images, list of images
+        shape0, shape1 = [], []  # image and inference shapes
+        for i, im in enumerate(imgs):
+            if isinstance(im, str):  # filename or uri
+                im = Image.open(requests.get(im, stream=True).raw if im.startswith('http') else im)  # open
+            im = np.array(im)  # to numpy
+            if im.shape[0] < 5:  # image in CHW
+                im = im.transpose((1, 2, 0))  # reverse dataloader .transpose(2, 0, 1)
+            im = im[:, :, :3] if im.ndim == 3 else np.tile(im[:, :, None], 3)  # enforce 3ch input
+            s = im.shape[:2]  # HWC
+            shape0.append(s)  # image shape
+            g = (size / max(s))  # gain
+            shape1.append([y * g for y in s])
+            imgs[i] = im  # update
+        shape1 = [make_divisible(x, int(self.stride.max())) for x in np.stack(shape1, 0).max(0)]  # inference shape
+        x = [letterbox(im, new_shape=shape1, auto=False)[0] for im in imgs]  # pad
+        x = np.stack(x, 0) if n > 1 else x[0][None]  # stack
+        x = np.ascontiguousarray(x.transpose((0, 3, 1, 2)))  # BHWC to BCHW
+        x = torch.from_numpy(x).to(p.device).type_as(p) / 255.  # uint8 to fp16/32
+
+        # Inference
+        with torch.no_grad():
+            y = self.model(x, augment, profile)[0]  # forward
+        y = non_max_suppression(y, conf_thres=self.conf, iou_thres=self.iou, classes=self.classes)  # NMS
+
+        # Post-process
+        for i in range(n):
+            scale_coords(shape1, y[i][:, :4], shape0[i])
+
+        return Detections(imgs, y, self.names)
+
+
+class Detections:
+    # detections class for YOLOv5 inference results
+    def __init__(self, imgs, pred, names=None):
+        super(Detections, self).__init__()
+        d = pred[0].device  # device
+        gn = [torch.tensor([*[im.shape[i] for i in [1, 0, 1, 0]], 1., 1.], device=d) for im in imgs]  # normalizations
+        self.imgs = imgs  # list of images as numpy arrays
+        self.pred = pred  # list of tensors pred[0] = (xyxy, conf, cls)
+        self.names = names  # class names
+        self.xyxy = pred  # xyxy pixels
+        self.xywh = [xyxy2xywh(x) for x in pred]  # xywh pixels
+        self.xyxyn = [x / g for x, g in zip(self.xyxy, gn)]  # xyxy normalized
+        self.xywhn = [x / g for x, g in zip(self.xywh, gn)]  # xywh normalized
+        self.n = len(self.pred)
+
+    def display(self, pprint=False, show=False, save=False, render=False):
+        colors = color_list()
+        for i, (img, pred) in enumerate(zip(self.imgs, self.pred)):
+            str = f'Image {i + 1}/{len(self.pred)}: {img.shape[0]}x{img.shape[1]} '
+            if pred is not None:
+                for c in pred[:, -1].unique():
+                    n = (pred[:, -1] == c).sum()  # detections per class
+                    str += f'{n} {self.names[int(c)]}s, '  # add to string
+                if show or save or render:
+                    img = Image.fromarray(img.astype(np.uint8)) if isinstance(img, np.ndarray) else img  # from np
+                    for *box, conf, cls in pred:  # xyxy, confidence, class
+                        # str += '%s %.2f, ' % (names[int(cls)], conf)  # label
+                        ImageDraw.Draw(img).rectangle(box, width=4, outline=colors[int(cls) % 10])  # plot
+            if pprint:
+                print(str)
+            if show:
+                img.show(f'Image {i}')  # show
+            if save:
+                f = f'results{i}.jpg'
+                str += f"saved to '{f}'"
+                img.save(f)  # save
+            if render:
+                self.imgs[i] = np.asarray(img)
+
+    def print(self):
+        self.display(pprint=True)  # print results
+
+    def show(self):
+        self.display(show=True)  # show results
+
+    def save(self):
+        self.display(save=True)  # save results
+
+    def render(self):
+        self.display(render=True)  # render results
+        return self.imgs
+
+    def __len__(self):
+        return self.n
+
+    def tolist(self):
+        # return a list of Detections objects, i.e. 'for result in results.tolist():'
+        x = [Detections([self.imgs[i]], [self.pred[i]], self.names) for i in range(self.n)]
+        for d in x:
+            for k in ['imgs', 'pred', 'xyxy', 'xyxyn', 'xywh', 'xywhn']:
+                setattr(d, k, getattr(d, k)[0])  # pop out of list
+        return x
+
+
+class Classify(nn.Module):
+    # Classification head, i.e. x(b,c1,20,20) to x(b,c2)
+    def __init__(self, c1, c2, k=1, s=1, p=None, g=1):  # ch_in, ch_out, kernel, stride, padding, groups
+        super(Classify, self).__init__()
+        self.aap = nn.AdaptiveAvgPool2d(1)  # to x(b,c1,1,1)
+        self.conv = nn.Conv2d(c1, c2, k, s, autopad(k, p), groups=g)  # to x(b,c2,1,1)
+        self.flat = nn.Flatten()
+
+    def forward(self, x):
+        z = torch.cat([self.aap(y) for y in (x if isinstance(x, list) else [x])], 1)  # cat if list
+        return self.flat(self.conv(z))  # flatten to x(b,c2)

models/experimental.py

@@ -0,0 +1,133 @@
+# This file contains experimental modules
+
+import numpy as np
+import torch
+import torch.nn as nn
+
+from models.common import Conv, DWConv
+from utils.google_utils import attempt_download
+
+
+class CrossConv(nn.Module):
+    # Cross Convolution Downsample
+    def __init__(self, c1, c2, k=3, s=1, g=1, e=1.0, shortcut=False):
+        # ch_in, ch_out, kernel, stride, groups, expansion, shortcut
+        super(CrossConv, self).__init__()
+        c_ = int(c2 * e)  # hidden channels
+        self.cv1 = Conv(c1, c_, (1, k), (1, s))
+        self.cv2 = Conv(c_, c2, (k, 1), (s, 1), g=g)
+        self.add = shortcut and c1 == c2
+
+    def forward(self, x):
+        return x + self.cv2(self.cv1(x)) if self.add else self.cv2(self.cv1(x))
+
+
+class Sum(nn.Module):
+    # Weighted sum of 2 or more layers https://arxiv.org/abs/1911.09070
+    def __init__(self, n, weight=False):  # n: number of inputs
+        super(Sum, self).__init__()
+        self.weight = weight  # apply weights boolean
+        self.iter = range(n - 1)  # iter object
+        if weight:
+            self.w = nn.Parameter(-torch.arange(1., n) / 2, requires_grad=True)  # layer weights
+
+    def forward(self, x):
+        y = x[0]  # no weight
+        if self.weight:
+            w = torch.sigmoid(self.w) * 2
+            for i in self.iter:
+                y = y + x[i + 1] * w[i]
+        else:
+            for i in self.iter:
+                y = y + x[i + 1]
+        return y
+
+
+class GhostConv(nn.Module):
+    # Ghost Convolution https://github.com/huawei-noah/ghostnet
+    def __init__(self, c1, c2, k=1, s=1, g=1, act=True):  # ch_in, ch_out, kernel, stride, groups
+        super(GhostConv, self).__init__()
+        c_ = c2 // 2  # hidden channels
+        self.cv1 = Conv(c1, c_, k, s, None, g, act)
+        self.cv2 = Conv(c_, c_, 5, 1, None, c_, act)
+
+    def forward(self, x):
+        y = self.cv1(x)
+        return torch.cat([y, self.cv2(y)], 1)
+
+
+class GhostBottleneck(nn.Module):
+    # Ghost Bottleneck https://github.com/huawei-noah/ghostnet
+    def __init__(self, c1, c2, k, s):
+        super(GhostBottleneck, self).__init__()
+        c_ = c2 // 2
+        self.conv = nn.Sequential(GhostConv(c1, c_, 1, 1),  # pw
+                                  DWConv(c_, c_, k, s, act=False) if s == 2 else nn.Identity(),  # dw
+                                  GhostConv(c_, c2, 1, 1, act=False))  # pw-linear
+        self.shortcut = nn.Sequential(DWConv(c1, c1, k, s, act=False),
+                                      Conv(c1, c2, 1, 1, act=False)) if s == 2 else nn.Identity()
+
+    def forward(self, x):
+        return self.conv(x) + self.shortcut(x)
+
+
+class MixConv2d(nn.Module):
+    # Mixed Depthwise Conv https://arxiv.org/abs/1907.09595
+    def __init__(self, c1, c2, k=(1, 3), s=1, equal_ch=True):
+        super(MixConv2d, self).__init__()
+        groups = len(k)
+        if equal_ch:  # equal c_ per group
+            i = torch.linspace(0, groups - 1E-6, c2).floor()  # c2 indices
+            c_ = [(i == g).sum() for g in range(groups)]  # intermediate channels
+        else:  # equal weight.numel() per group
+            b = [c2] + [0] * groups
+            a = np.eye(groups + 1, groups, k=-1)
+            a -= np.roll(a, 1, axis=1)
+            a *= np.array(k) ** 2
+            a[0] = 1
+            c_ = np.linalg.lstsq(a, b, rcond=None)[0].round()  # solve for equal weight indices, ax = b
+
+        self.m = nn.ModuleList([nn.Conv2d(c1, int(c_[g]), k[g], s, k[g] // 2, bias=False) for g in range(groups)])
+        self.bn = nn.BatchNorm2d(c2)
+        self.act = nn.LeakyReLU(0.1, inplace=True)
+
+    def forward(self, x):
+        return x + self.act(self.bn(torch.cat([m(x) for m in self.m], 1)))
+
+
+class Ensemble(nn.ModuleList):
+    # Ensemble of models
+    def __init__(self):
+        super(Ensemble, self).__init__()
+
+    def forward(self, x, augment=False):
+        y = []
+        for module in self:
+            y.append(module(x, augment)[0])
+        # y = torch.stack(y).max(0)[0]  # max ensemble
+        # y = torch.stack(y).mean(0)  # mean ensemble
+        y = torch.cat(y, 1)  # nms ensemble
+        return y, None  # inference, train output
+
+
+def attempt_load(weights, map_location=None):
+    # Loads an ensemble of models weights=[a,b,c] or a single model weights=[a] or weights=a
+    model = Ensemble()
+    for w in weights if isinstance(weights, list) else [weights]:
+        attempt_download(w)
+        model.append(torch.load(w, map_location=map_location)['model'].float().fuse().eval())  # load FP32 model
+
+    # Compatibility updates
+    for m in model.modules():
+        if type(m) in [nn.Hardswish, nn.LeakyReLU, nn.ReLU, nn.ReLU6, nn.SiLU]:
+            m.inplace = True  # pytorch 1.7.0 compatibility
+        elif type(m) is Conv:
+            m._non_persistent_buffers_set = set()  # pytorch 1.6.0 compatibility
+
+    if len(model) == 1:
+        return model[-1]  # return model
+    else:
+        print('Ensemble created with %s\n' % weights)
+        for k in ['names', 'stride']:
+            setattr(model, k, getattr(model[-1], k))
+        return model  # return ensemble

models/yolo.py

@@ -0,0 +1,350 @@
+import argparse
+import logging
+import math
+import sys
+from copy import deepcopy
+from pathlib import Path
+
+import torch
+import torch.nn as nn
+
+sys.path.append('./')  # to run '$ python *.py' files in subdirectories
+logger = logging.getLogger(__name__)
+
+from models.common import Conv, Bottleneck, SPP, DWConv, Focus, BottleneckCSP, C3, ShuffleV2Block, Concat, NMS, autoShape, StemBlock, BlazeBlock, DoubleBlazeBlock
+from models.experimental import MixConv2d, CrossConv
+from utils.autoanchor import check_anchor_order
+from utils.general import make_divisible, check_file, set_logging
+from utils.torch_utils import time_synchronized, fuse_conv_and_bn, model_info, scale_img, initialize_weights, \
+    select_device, copy_attr
+
+try:
+    import thop  # for FLOPS computation
+except ImportError:
+    thop = None
+
+
+class Detect(nn.Module):
+    stride = None  # strides computed during build
+    export_cat = False  # onnx export cat output
+
+    def __init__(self, nc=80, anchors=(), ch=()):  # detection layer
+        super(Detect, self).__init__()
+        self.nc = nc  # number of classes
+        #self.no = nc + 5  # number of outputs per anchor
+        self.no = nc + 5 + 8  # number of outputs per anchor
+
+        self.nl = len(anchors)  # number of detection layers
+        self.na = len(anchors[0]) // 2  # number of anchors
+        self.grid = [torch.zeros(1)] * self.nl  # init grid
+        a = torch.tensor(anchors).float().view(self.nl, -1, 2)
+        self.register_buffer('anchors', a)  # shape(nl,na,2)
+        self.register_buffer('anchor_grid', a.clone().view(self.nl, 1, -1, 1, 1, 2))  # shape(nl,1,na,1,1,2)
+        self.m = nn.ModuleList(nn.Conv2d(x, self.no * self.na, 1) for x in ch)  # output conv
+
+    def forward(self, x):
+        # x = x.copy()  # for profiling
+        z = []  # inference output
+        if self.export_cat:
+            for i in range(self.nl):
+                x[i] = self.m[i](x[i])  # conv
+                bs, _, ny, nx = x[i].shape  # x(bs,255,20,20) to x(bs,3,20,20,85)
+                x[i] = x[i].view(bs, self.na, self.no, ny, nx).permute(0, 1, 3, 4, 2).contiguous()
+
+                if self.grid[i].shape[2:4] != x[i].shape[2:4]:
+                    # self.grid[i] = self._make_grid(nx, ny).to(x[i].device)
+                    self.grid[i], self.anchor_grid[i] = self._make_grid_new(nx, ny,i)
+
+                y = torch.full_like(x[i], 0)
+                y = y + torch.cat((x[i][:, :, :, :, 0:5].sigmoid(), torch.cat((x[i][:, :, :, :, 5:13], x[i][:, :, :, :, 13:13+self.nc].sigmoid()), 4)), 4)
+
+                box_xy = (y[:, :, :, :, 0:2] * 2. - 0.5 + self.grid[i].to(x[i].device)) * self.stride[i] # xy
+                box_wh = (y[:, :, :, :, 2:4] * 2) ** 2 * self.anchor_grid[i] # wh
+                # box_conf = torch.cat((box_xy, torch.cat((box_wh, y[:, :, :, :, 4:5]), 4)), 4)
+
+                landm1 = y[:, :, :, :, 5:7] * self.anchor_grid[i] + self.grid[i].to(x[i].device) * self.stride[i]  # landmark x1 y1
+                landm2 = y[:, :, :, :, 7:9] * self.anchor_grid[i] + self.grid[i].to(x[i].device) * self.stride[i]  # landmark x2 y2
+                landm3 = y[:, :, :, :, 9:11] * self.anchor_grid[i] + self.grid[i].to(x[i].device) * self.stride[i]  # landmark x3 y3
+                landm4 = y[:, :, :, :, 11:13] * self.anchor_grid[i] + self.grid[i].to(x[i].device) * self.stride[i]  # landmark x4 y4
+                prob=    y[:, :, :, :, 13:13+self.nc]
+                score,index_ = torch.max(prob,dim=-1,keepdim=True)
+                score=score.type(box_xy.dtype)
+                index_=index_.type(box_xy.dtype)
+                index =torch.argmax(prob,dim=-1,keepdim=True).type(box_xy.dtype)
+                # landm5 = y[:, :, :, :, 13:13] * self.anchor_grid[i] + self.grid[i].to(x[i].device) * self.stride[i]  # landmark x5 y5
+                # landm = torch.cat((landm1, torch.cat((landm2, torch.cat((landm3, torch.cat((landm4, landm5), 4)), 4)), 4)), 4)
+                # y = torch.cat((box_conf, torch.cat((landm, y[:, :, :, :, 13:13+self.nc]), 4)), 4)
+                y = torch.cat([box_xy, box_wh, y[:, :, :, :, 4:5], landm1, landm2, landm3, landm4, y[:, :, :, :, 13:13+self.nc]], -1)
+
+                z.append(y.view(bs, -1, self.no))
+            return torch.cat(z, 1)
+        
+        for i in range(self.nl):
+            x[i] = self.m[i](x[i])  # conv
+            bs, _, ny, nx = x[i].shape  # x(bs,255,20,20) to x(bs,3,20,20,85)
+            x[i] = x[i].view(bs, self.na, self.no, ny, nx).permute(0, 1, 3, 4, 2).contiguous()
+
+            if not self.training:  # inference
+                if self.grid[i].shape[2:4] != x[i].shape[2:4]:
+                    self.grid[i] = self._make_grid(nx, ny).to(x[i].device)
+
+                y = torch.full_like(x[i], 0)
+                class_range = list(range(5)) + list(range(13,13+self.nc))
+                y[..., class_range] = x[i][..., class_range].sigmoid()
+                y[..., 5:13] = x[i][..., 5:13]
+                #y = x[i].sigmoid()
+
+                y[..., 0:2] = (y[..., 0:2] * 2. - 0.5 + self.grid[i].to(x[i].device)) * self.stride[i]  # xy
+                y[..., 2:4] = (y[..., 2:4] * 2) ** 2 * self.anchor_grid[i]  # wh
+
+                #y[..., 5:13] = y[..., 5:13] * 8 - 4
+                y[..., 5:7]   = y[..., 5:7] *   self.anchor_grid[i] + self.grid[i].to(x[i].device) * self.stride[i] # landmark x1 y1
+                y[..., 7:9]   = y[..., 7:9] *   self.anchor_grid[i] + self.grid[i].to(x[i].device) * self.stride[i]# landmark x2 y2
+                y[..., 9:11]  = y[..., 9:11] *  self.anchor_grid[i] + self.grid[i].to(x[i].device) * self.stride[i]# landmark x3 y3
+                y[..., 11:13] = y[..., 11:13] * self.anchor_grid[i] + self.grid[i].to(x[i].device) * self.stride[i]# landmark x4 y4
+                # y[..., 13:13] = y[..., 13:13] * self.anchor_grid[i] + self.grid[i].to(x[i].device) * self.stride[i]# landmark x5 y5
+
+                #y[..., 5:7] = (y[..., 5:7] * 2 -1) * self.anchor_grid[i]  # landmark x1 y1
+                #y[..., 7:9] = (y[..., 7:9] * 2 -1) * self.anchor_grid[i]  # landmark x2 y2
+                #y[..., 9:11] = (y[..., 9:11] * 2 -1) * self.anchor_grid[i]  # landmark x3 y3
+                #y[..., 11:13] = (y[..., 11:13] * 2 -1) * self.anchor_grid[i]  # landmark x4 y4
+                #y[..., 13:13] = (y[..., 13:13] * 2 -1) * self.anchor_grid[i]  # landmark x5 y5
+
+                z.append(y.view(bs, -1, self.no))
+
+        return x if self.training else (torch.cat(z, 1), x)
+
+    @staticmethod
+    def _make_grid(nx=20, ny=20):
+        yv, xv = torch.meshgrid([torch.arange(ny), torch.arange(nx)])
+        return torch.stack((xv, yv), 2).view((1, 1, ny, nx, 2)).float()
+
+    def _make_grid_new(self,nx=20, ny=20,i=0):
+        d = self.anchors[i].device
+        if '1.10.0' in torch.__version__: # torch>=1.10.0 meshgrid workaround for torch>=0.7 compatibility
+            yv, xv = torch.meshgrid([torch.arange(ny).to(d), torch.arange(nx).to(d)], indexing='ij')
+        else:
+            yv, xv = torch.meshgrid([torch.arange(ny).to(d), torch.arange(nx).to(d)])
+        grid = torch.stack((xv, yv), 2).expand((1, self.na, ny, nx, 2)).float()
+        anchor_grid = (self.anchors[i].clone() * self.stride[i]).view((1, self.na, 1, 1, 2)).expand((1, self.na, ny, nx, 2)).float()
+        return grid, anchor_grid
+class Model(nn.Module):
+    def __init__(self, cfg='yolov5s.yaml', ch=3, nc=None):  # model, input channels, number of classes
+        super(Model, self).__init__()
+        if isinstance(cfg, dict):
+            self.yaml = cfg  # model dict
+        else:  # is *.yaml
+            import yaml  # for torch hub
+            self.yaml_file = Path(cfg).name
+            with open(cfg) as f:
+                self.yaml = yaml.load(f, Loader=yaml.FullLoader)  # model dict
+
+        # Define model
+        ch = self.yaml['ch'] = self.yaml.get('ch', ch)  # input channels
+        if nc and nc != self.yaml['nc']:
+            logger.info('Overriding model.yaml nc=%g with nc=%g' % (self.yaml['nc'], nc))
+            self.yaml['nc'] = nc  # override yaml value
+        self.model, self.save = parse_model(deepcopy(self.yaml), ch=[ch])  # model, savelist
+        self.names = [str(i) for i in range(self.yaml['nc'])]  # default names
+        # print([x.shape for x in self.forward(torch.zeros(1, ch, 64, 64))])
+
+        # Build strides, anchors
+        m = self.model[-1]  # Detect()
+        if isinstance(m, Detect):
+            s = 128  # 2x min stride
+            m.stride = torch.tensor([s / x.shape[-2] for x in self.forward(torch.zeros(1, ch, s, s))])  # forward
+            m.anchors /= m.stride.view(-1, 1, 1)
+            check_anchor_order(m)
+            self.stride = m.stride
+            self._initialize_biases()  # only run once
+            # print('Strides: %s' % m.stride.tolist())
+
+        # Init weights, biases
+        initialize_weights(self)
+        self.info()
+        logger.info('')
+
+    def forward(self, x, augment=False, profile=False):
+        if augment:
+            img_size = x.shape[-2:]  # height, width
+            s = [1, 0.83, 0.67]  # scales
+            f = [None, 3, None]  # flips (2-ud, 3-lr)
+            y = []  # outputs
+            for si, fi in zip(s, f):
+                xi = scale_img(x.flip(fi) if fi else x, si)
+                yi = self.forward_once(xi)[0]  # forward
+                # cv2.imwrite('img%g.jpg' % s, 255 * xi[0].numpy().transpose((1, 2, 0))[:, :, ::-1])  # save
+                yi[..., :4] /= si  # de-scale
+                if fi == 2:
+                    yi[..., 1] = img_size[0] - yi[..., 1]  # de-flip ud
+                elif fi == 3:
+                    yi[..., 0] = img_size[1] - yi[..., 0]  # de-flip lr
+                y.append(yi)
+            return torch.cat(y, 1), None  # augmented inference, train
+        else:
+            return self.forward_once(x, profile)  # single-scale inference, train
+
+    def forward_once(self, x, profile=False):
+        y, dt = [], []  # outputs
+        for m in self.model:
+            if m.f != -1:  # if not from previous layer
+                x = y[m.f] if isinstance(m.f, int) else [x if j == -1 else y[j] for j in m.f]  # from earlier layers
+
+            if profile:
+                o = thop.profile(m, inputs=(x,), verbose=False)[0] / 1E9 * 2 if thop else 0  # FLOPS
+                t = time_synchronized()
+                for _ in range(10):
+                    _ = m(x)
+                dt.append((time_synchronized() - t) * 100)
+                print('%10.1f%10.0f%10.1fms %-40s' % (o, m.np, dt[-1], m.type))
+
+            x = m(x)  # run
+            y.append(x if m.i in self.save else None)  # save output
+
+        if profile:
+            print('%.1fms total' % sum(dt))
+        return x
+
+    def _initialize_biases(self, cf=None):  # initialize biases into Detect(), cf is class frequency
+        # https://arxiv.org/abs/1708.02002 section 3.3
+        # cf = torch.bincount(torch.tensor(np.concatenate(dataset.labels, 0)[:, 0]).long(), minlength=nc) + 1.
+        m = self.model[-1]  # Detect() module
+        for mi, s in zip(m.m, m.stride):  # from
+            b = mi.bias.view(m.na, -1)  # conv.bias(255) to (3,85)
+            b.data[:, 4] += math.log(8 / (640 / s) ** 2)  # obj (8 objects per 640 image)
+            b.data[:, 5:] += math.log(0.6 / (m.nc - 0.99)) if cf is None else torch.log(cf / cf.sum())  # cls
+            mi.bias = torch.nn.Parameter(b.view(-1), requires_grad=True)
+
+    def _print_biases(self):
+        m = self.model[-1]  # Detect() module
+        for mi in m.m:  # from
+            b = mi.bias.detach().view(m.na, -1).T  # conv.bias(255) to (3,85)
+            print(('%6g Conv2d.bias:' + '%10.3g' * 6) % (mi.weight.shape[1], *b[:5].mean(1).tolist(), b[5:].mean()))
+
+    # def _print_weights(self):
+    #     for m in self.model.modules():
+    #         if type(m) is Bottleneck:
+    #             print('%10.3g' % (m.w.detach().sigmoid() * 2))  # shortcut weights
+
+    def fuse(self):  # fuse model Conv2d() + BatchNorm2d() layers
+        print('Fusing layers... ')
+        for m in self.model.modules():
+            if type(m) is Conv and hasattr(m, 'bn'):
+                m.conv = fuse_conv_and_bn(m.conv, m.bn)  # update conv
+                delattr(m, 'bn')  # remove batchnorm
+                m.forward = m.fuseforward  # update forward
+            elif type(m) is nn.Upsample:
+                m.recompute_scale_factor = None  # torch 1.11.0 compatibility
+        self.info()
+        return self
+
+    def nms(self, mode=True):  # add or remove NMS module
+        present = type(self.model[-1]) is NMS  # last layer is NMS
+        if mode and not present:
+            print('Adding NMS... ')
+            m = NMS()  # module
+            m.f = -1  # from
+            m.i = self.model[-1].i + 1  # index
+            self.model.add_module(name='%s' % m.i, module=m)  # add
+            self.eval()
+        elif not mode and present:
+            print('Removing NMS... ')
+            self.model = self.model[:-1]  # remove
+        return self
+
+    def autoshape(self):  # add autoShape module
+        print('Adding autoShape... ')
+        m = autoShape(self)  # wrap model
+        copy_attr(m, self, include=('yaml', 'nc', 'hyp', 'names', 'stride'), exclude=())  # copy attributes
+        return m
+
+    def info(self, verbose=False, img_size=640):  # print model information
+        model_info(self, verbose, img_size)
+
+
+def parse_model(d, ch):  # model_dict, input_channels(3)
+    logger.info('\n%3s%18s%3s%10s  %-40s%-30s' % ('', 'from', 'n', 'params', 'module', 'arguments'))
+    anchors, nc, gd, gw = d['anchors'], d['nc'], d['depth_multiple'], d['width_multiple']
+    na = (len(anchors[0]) // 2) if isinstance(anchors, list) else anchors  # number of anchors
+    no = na * (nc + 5)  # number of outputs = anchors * (classes + 5)
+
+    layers, save, c2 = [], [], ch[-1]  # layers, savelist, ch out
+    for i, (f, n, m, args) in enumerate(d['backbone'] + d['head']):  # from, number, module, args
+        m = eval(m) if isinstance(m, str) else m  # eval strings
+        for j, a in enumerate(args):
+            try:
+                args[j] = eval(a) if isinstance(a, str) else a  # eval strings
+            except:
+                pass
+
+        n = max(round(n * gd), 1) if n > 1 else n  # depth gain
+        if m in [Conv, Bottleneck, SPP, DWConv, MixConv2d, Focus, CrossConv, BottleneckCSP, C3, ShuffleV2Block, StemBlock, BlazeBlock, DoubleBlazeBlock]:
+            c1, c2 = ch[f], args[0]
+
+            # Normal
+            # if i > 0 and args[0] != no:  # channel expansion factor
+            #     ex = 1.75  # exponential (default 2.0)
+            #     e = math.log(c2 / ch[1]) / math.log(2)
+            #     c2 = int(ch[1] * ex ** e)
+            # if m != Focus:
+
+            c2 = make_divisible(c2 * gw, 8) if c2 != no else c2
+
+            # Experimental
+            # if i > 0 and args[0] != no:  # channel expansion factor
+            #     ex = 1 + gw  # exponential (default 2.0)
+            #     ch1 = 32  # ch[1]
+            #     e = math.log(c2 / ch1) / math.log(2)  # level 1-n
+            #     c2 = int(ch1 * ex ** e)
+            # if m != Focus:
+            #     c2 = make_divisible(c2, 8) if c2 != no else c2
+
+            args = [c1, c2, *args[1:]]
+            if m in [BottleneckCSP, C3]:
+                args.insert(2, n)
+                n = 1
+        elif m is nn.BatchNorm2d:
+            args = [ch[f]]
+        elif m is Concat:
+            c2 = sum([ch[-1 if x == -1 else x + 1] for x in f])
+        elif m is Detect:
+            args.append([ch[x + 1] for x in f])
+            if isinstance(args[1], int):  # number of anchors
+                args[1] = [list(range(args[1] * 2))] * len(f)
+        else:
+            c2 = ch[f]
+
+        m_ = nn.Sequential(*[m(*args) for _ in range(n)]) if n > 1 else m(*args)  # module
+        t = str(m)[8:-2].replace('__main__.', '')  # module type
+        np = sum([x.numel() for x in m_.parameters()])  # number params
+        m_.i, m_.f, m_.type, m_.np = i, f, t, np  # attach index, 'from' index, type, number params
+        logger.info('%3s%18s%3s%10.0f  %-40s%-30s' % (i, f, n, np, t, args))  # print
+        save.extend(x % i for x in ([f] if isinstance(f, int) else f) if x != -1)  # append to savelist
+        layers.append(m_)
+        ch.append(c2)
+    return nn.Sequential(*layers), sorted(save)
+
+
+from thop import profile
+from thop import clever_format
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--cfg', type=str, default='yolov5s.yaml', help='model.yaml')
+    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    opt = parser.parse_args()
+    opt.cfg = check_file(opt.cfg)  # check file
+    set_logging()
+    device = select_device(opt.device)
+    
+    # Create model
+    model = Model(opt.cfg).to(device)
+    stride = model.stride.max()
+    if stride == 32:
+        input = torch.Tensor(1, 3, 480, 640).to(device)
+    else:
+        input = torch.Tensor(1, 3, 512, 640).to(device)
+    model.train()
+    print(model)
+    flops, params = profile(model, inputs=(input, ))
+    flops, params = clever_format([flops, params], "%.3f")
+    print('Flops:', flops, ',Params:' ,params)

models/yolov5l.yaml

@@ -0,0 +1,47 @@
+# parameters
+nc: 1  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+
+# anchors
+anchors:
+  - [4,5,  8,10,  13,16]  # P3/8
+  - [23,29,  43,55,  73,105]  # P4/16
+  - [146,217,  231,300,  335,433]  # P5/32
+
+# YOLOv5 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, StemBlock, [64, 3, 2]],  # 0-P1/2
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],      # 2-P3/8
+   [-1, 9, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],      # 4-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],     # 6-P5/32
+   [-1, 1, SPP, [1024, [3,5,7]]],
+   [-1, 3, C3, [1024, False]],      # 8
+  ]
+
+# YOLOv5 head
+head:
+  [[-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 5], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 12
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 3], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 16 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 13], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 19 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 9], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [1024, False]],  # 22 (P5/32-large)
+
+   [[16, 19, 22], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]

models/yolov5l6.yaml

@@ -0,0 +1,60 @@
+# parameters
+nc: 1  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+
+# anchors
+anchors:
+  - [6,7,  9,11,  13,16]  # P3/8
+  - [18,23,  26,33,  37,47]  # P4/16
+  - [54,67,  77,104,  112,154]  # P5/32
+  - [174,238,  258,355,  445,568]  # P6/64
+
+# YOLOv5 backbone
+backbone:
+  # [from, number, module, args]
+  [ [ -1, 1, StemBlock, [ 64, 3, 2 ] ],  # 0-P1/2
+    [ -1, 3, C3, [ 128 ] ],
+    [ -1, 1, Conv, [ 256, 3, 2 ] ],  # 2-P3/8
+    [ -1, 9, C3, [ 256 ] ],
+    [ -1, 1, Conv, [ 512, 3, 2 ] ],  # 4-P4/16
+    [ -1, 9, C3, [ 512 ] ],
+    [ -1, 1, Conv, [ 768, 3, 2 ] ],  # 6-P5/32
+    [ -1, 3, C3, [ 768 ] ],
+    [ -1, 1, Conv, [ 1024, 3, 2 ] ],  # 8-P6/64
+    [ -1, 1, SPP, [ 1024, [ 3, 5, 7 ] ] ],
+    [ -1, 3, C3, [ 1024, False ] ],  # 10
+  ]
+
+# YOLOv5 head
+head:
+  [ [ -1, 1, Conv, [ 768, 1, 1 ] ],
+    [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
+    [ [ -1, 7 ], 1, Concat, [ 1 ] ],  # cat backbone P5
+    [ -1, 3, C3, [ 768, False ] ],  # 14
+
+    [ -1, 1, Conv, [ 512, 1, 1 ] ],
+    [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
+    [ [ -1, 5 ], 1, Concat, [ 1 ] ],  # cat backbone P4
+    [ -1, 3, C3, [ 512, False ] ],  # 18
+
+    [ -1, 1, Conv, [ 256, 1, 1 ] ],
+    [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
+    [ [ -1, 3 ], 1, Concat, [ 1 ] ],  # cat backbone P3
+    [ -1, 3, C3, [ 256, False ] ],  # 22 (P3/8-small)
+
+    [ -1, 1, Conv, [ 256, 3, 2 ] ],
+    [ [ -1, 19 ], 1, Concat, [ 1 ] ],  # cat head P4
+    [ -1, 3, C3, [ 512, False ] ],  # 25 (P4/16-medium)
+
+    [ -1, 1, Conv, [ 512, 3, 2 ] ],
+    [ [ -1, 15 ], 1, Concat, [ 1 ] ],  # cat head P5
+    [ -1, 3, C3, [ 768, False ] ],  # 28 (P5/32-large)
+
+    [ -1, 1, Conv, [ 768, 3, 2 ] ],
+    [ [ -1, 11 ], 1, Concat, [ 1 ] ],  # cat head P6
+    [ -1, 3, C3, [ 1024, False ] ],  # 31 (P6/64-xlarge)
+
+    [ [ 22, 25, 28, 31 ], 1, Detect, [ nc, anchors ] ],  # Detect(P3, P4, P5, P6)
+  ]
+

models/yolov5m.yaml

@@ -0,0 +1,47 @@
+# parameters
+nc: 1  # number of classes
+depth_multiple: 0.67  # model depth multiple
+width_multiple: 0.75  # layer channel multiple
+
+# anchors
+anchors:
+  - [4,5,  8,10,  13,16]  # P3/8
+  - [23,29,  43,55,  73,105]  # P4/16
+  - [146,217,  231,300,  335,433]  # P5/32
+
+# YOLOv5 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, StemBlock, [64, 3, 2]],  # 0-P1/2
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],      # 2-P3/8
+   [-1, 9, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],      # 4-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],     # 6-P5/32
+   [-1, 1, SPP, [1024, [3,5,7]]],
+   [-1, 3, C3, [1024, False]],      # 8
+  ]
+
+# YOLOv5 head
+head:
+  [[-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 5], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 12
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 3], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 16 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 13], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 19 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 9], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [1024, False]],  # 22 (P5/32-large)
+
+   [[16, 19, 22], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]

models/yolov5m6.yaml

@@ -0,0 +1,60 @@
+# parameters
+nc: 1  # number of classes
+depth_multiple: 0.67  # model depth multiple
+width_multiple: 0.75  # layer channel multiple
+
+# anchors
+anchors:
+  - [6,7,  9,11,  13,16]  # P3/8
+  - [18,23,  26,33,  37,47]  # P4/16
+  - [54,67,  77,104,  112,154]  # P5/32
+  - [174,238,  258,355,  445,568]  # P6/64
+
+# YOLOv5 backbone
+backbone:
+  # [from, number, module, args]
+  [ [ -1, 1, StemBlock, [ 64, 3, 2 ] ],  # 0-P1/2
+    [ -1, 3, C3, [ 128 ] ],
+    [ -1, 1, Conv, [ 256, 3, 2 ] ],  # 2-P3/8
+    [ -1, 9, C3, [ 256 ] ],
+    [ -1, 1, Conv, [ 512, 3, 2 ] ],  # 4-P4/16
+    [ -1, 9, C3, [ 512 ] ],
+    [ -1, 1, Conv, [ 768, 3, 2 ] ],  # 6-P5/32
+    [ -1, 3, C3, [ 768 ] ],
+    [ -1, 1, Conv, [ 1024, 3, 2 ] ],  # 8-P6/64
+    [ -1, 1, SPP, [ 1024, [ 3, 5, 7 ] ] ],
+    [ -1, 3, C3, [ 1024, False ] ],  # 10
+  ]
+
+# YOLOv5 head
+head:
+  [ [ -1, 1, Conv, [ 768, 1, 1 ] ],
+    [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
+    [ [ -1, 7 ], 1, Concat, [ 1 ] ],  # cat backbone P5
+    [ -1, 3, C3, [ 768, False ] ],  # 14
+
+    [ -1, 1, Conv, [ 512, 1, 1 ] ],
+    [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
+    [ [ -1, 5 ], 1, Concat, [ 1 ] ],  # cat backbone P4
+    [ -1, 3, C3, [ 512, False ] ],  # 18
+
+    [ -1, 1, Conv, [ 256, 1, 1 ] ],
+    [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
+    [ [ -1, 3 ], 1, Concat, [ 1 ] ],  # cat backbone P3
+    [ -1, 3, C3, [ 256, False ] ],  # 22 (P3/8-small)
+
+    [ -1, 1, Conv, [ 256, 3, 2 ] ],
+    [ [ -1, 19 ], 1, Concat, [ 1 ] ],  # cat head P4
+    [ -1, 3, C3, [ 512, False ] ],  # 25 (P4/16-medium)
+
+    [ -1, 1, Conv, [ 512, 3, 2 ] ],
+    [ [ -1, 15 ], 1, Concat, [ 1 ] ],  # cat head P5
+    [ -1, 3, C3, [ 768, False ] ],  # 28 (P5/32-large)
+
+    [ -1, 1, Conv, [ 768, 3, 2 ] ],
+    [ [ -1, 11 ], 1, Concat, [ 1 ] ],  # cat head P6
+    [ -1, 3, C3, [ 1024, False ] ],  # 31 (P6/64-xlarge)
+
+    [ [ 22, 25, 28, 31 ], 1, Detect, [ nc, anchors ] ],  # Detect(P3, P4, P5, P6)
+  ]
+

models/yolov5n-0.5.yaml

@@ -0,0 +1,46 @@
+# parameters
+nc: 1  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 0.5  # layer channel multiple
+
+# anchors
+anchors:
+  - [4,5,  8,10,  13,16]  # P3/8
+  - [23,29,  43,55,  73,105]  # P4/16
+  - [146,217,  231,300,  335,433]  # P5/32
+
+# YOLOv5 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, StemBlock, [32, 3, 2]],    # 0-P2/4
+   [-1, 1, ShuffleV2Block, [128, 2]], # 1-P3/8
+   [-1, 3, ShuffleV2Block, [128, 1]], # 2
+   [-1, 1, ShuffleV2Block, [256, 2]], # 3-P4/16
+   [-1, 7, ShuffleV2Block, [256, 1]], # 4
+   [-1, 1, ShuffleV2Block, [512, 2]], # 5-P5/32
+   [-1, 3, ShuffleV2Block, [512, 1]], # 6
+  ]
+
+# YOLOv5 head
+head:
+  [[-1, 1, Conv, [128, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P4
+   [-1, 1, C3, [128, False]],  # 10
+
+   [-1, 1, Conv, [128, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 2], 1, Concat, [1]],  # cat backbone P3
+   [-1, 1, C3, [128, False]],  # 14 (P3/8-small)
+
+   [-1, 1, Conv, [128, 3, 2]],
+   [[-1, 11], 1, Concat, [1]],  # cat head P4
+   [-1, 1, C3, [128, False]],  # 17 (P4/16-medium)
+
+   [-1, 1, Conv, [128, 3, 2]],
+   [[-1, 7], 1, Concat, [1]],  # cat head P5
+   [-1, 1, C3, [128, False]],  # 20 (P5/32-large)
+
+   [[14, 17, 20], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]
+          

models/yolov5n.yaml

@@ -0,0 +1,46 @@
+# parameters
+nc: 1  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+
+# anchors
+anchors:
+  - [4,5,  8,10,  13,16]  # P3/8
+  - [23,29,  43,55,  73,105]  # P4/16
+  - [146,217,  231,300,  335,433]  # P5/32
+
+# YOLOv5 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, StemBlock, [32, 3, 2]],    # 0-P2/4
+   [-1, 1, ShuffleV2Block, [128, 2]], # 1-P3/8
+   [-1, 3, ShuffleV2Block, [128, 1]], # 2
+   [-1, 1, ShuffleV2Block, [256, 2]], # 3-P4/16
+   [-1, 7, ShuffleV2Block, [256, 1]], # 4
+   [-1, 1, ShuffleV2Block, [512, 2]], # 5-P5/32
+   [-1, 3, ShuffleV2Block, [512, 1]], # 6
+  ]
+
+# YOLOv5 head
+head:
+  [[-1, 1, Conv, [128, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P4
+   [-1, 1, C3, [128, False]],  # 10
+
+   [-1, 1, Conv, [128, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 2], 1, Concat, [1]],  # cat backbone P3
+   [-1, 1, C3, [128, False]],  # 14 (P3/8-small)
+
+   [-1, 1, Conv, [128, 3, 2]],
+   [[-1, 11], 1, Concat, [1]],  # cat head P4
+   [-1, 1, C3, [128, False]],  # 17 (P4/16-medium)
+
+   [-1, 1, Conv, [128, 3, 2]],
+   [[-1, 7], 1, Concat, [1]],  # cat head P5
+   [-1, 1, C3, [128, False]],  # 20 (P5/32-large)
+
+   [[14, 17, 20], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]
+          

models/yolov5n6.yaml

@@ -0,0 +1,58 @@
+# parameters
+nc: 1  # number of classes
+depth_multiple: 1.0  # model depth multiple
+width_multiple: 1.0  # layer channel multiple
+
+# anchors
+anchors:
+  - [6,7,  9,11,  13,16]  # P3/8
+  - [18,23,  26,33,  37,47]  # P4/16
+  - [54,67,  77,104,  112,154]  # P5/32
+  - [174,238,  258,355,  445,568]  # P6/64
+
+# YOLOv5 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, StemBlock, [32, 3, 2]],    # 0-P2/4
+   [-1, 1, ShuffleV2Block, [128, 2]], # 1-P3/8
+   [-1, 3, ShuffleV2Block, [128, 1]], # 2
+   [-1, 1, ShuffleV2Block, [256, 2]], # 3-P4/16
+   [-1, 7, ShuffleV2Block, [256, 1]], # 4
+   [-1, 1, ShuffleV2Block, [384, 2]], # 5-P5/32
+   [-1, 3, ShuffleV2Block, [384, 1]], # 6
+   [-1, 1, ShuffleV2Block, [512, 2]], # 7-P6/64
+   [-1, 3, ShuffleV2Block, [512, 1]], # 8
+  ]
+
+# YOLOv5 head
+head:
+  [[-1, 1, Conv, [128, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 6], 1, Concat, [1]],  # cat backbone P5
+   [-1, 1, C3, [128, False]],  # 12
+
+   [-1, 1, Conv, [128, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 4], 1, Concat, [1]],  # cat backbone P4
+   [-1, 1, C3, [128, False]],  # 16 (P4/8-small)
+
+   [-1, 1, Conv, [128, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 2], 1, Concat, [1]],  # cat backbone P3
+   [-1, 1, C3, [128, False]],  # 20 (P3/8-small)
+
+   [-1, 1, Conv, [128, 3, 2]],
+   [[-1, 17], 1, Concat, [1]],  # cat head P4
+   [-1, 1, C3, [128, False]],  # 23 (P4/16-medium)
+
+   [-1, 1, Conv, [128, 3, 2]],
+   [[-1, 13], 1, Concat, [1]],  # cat head P5
+   [-1, 1, C3, [128, False]],  # 26 (P5/32-large)
+
+   [-1, 1, Conv, [128, 3, 2]],
+   [[-1, 9], 1, Concat, [1]],  # cat head P6
+   [-1, 1, C3, [128, False]],  # 29 (P6/64-large)
+
+   [[20, 23, 26, 29], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]
+          

models/yolov5s.yaml

@@ -0,0 +1,47 @@
+# parameters
+nc: 1  # number of classes
+depth_multiple: 0.33  # model depth multiple
+width_multiple: 0.5  # layer channel multiple
+
+# anchors
+anchors:
+  - [4,5,  8,10,  13,16]  # P3/8
+  - [23,29,  43,55,  73,105]  # P4/16
+  - [146,217,  231,300,  335,433]  # P5/32
+
+# YOLOv5 backbone
+backbone:
+  # [from, number, module, args]
+  [[-1, 1, StemBlock, [64, 3, 2]],  # 0-P1/2
+   [-1, 3, C3, [128]],
+   [-1, 1, Conv, [256, 3, 2]],      # 2-P3/8
+   [-1, 9, C3, [256]],
+   [-1, 1, Conv, [512, 3, 2]],      # 4-P4/16
+   [-1, 9, C3, [512]],
+   [-1, 1, Conv, [1024, 3, 2]],     # 6-P5/32
+   [-1, 1, SPP, [1024, [3,5,7]]],
+   [-1, 3, C3, [1024, False]],      # 8
+  ]
+
+# YOLOv5 head
+head:
+  [[-1, 1, Conv, [512, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 5], 1, Concat, [1]],  # cat backbone P4
+   [-1, 3, C3, [512, False]],  # 12
+
+   [-1, 1, Conv, [256, 1, 1]],
+   [-1, 1, nn.Upsample, [None, 2, 'nearest']],
+   [[-1, 3], 1, Concat, [1]],  # cat backbone P3
+   [-1, 3, C3, [256, False]],  # 16 (P3/8-small)
+
+   [-1, 1, Conv, [256, 3, 2]],
+   [[-1, 13], 1, Concat, [1]],  # cat head P4
+   [-1, 3, C3, [512, False]],  # 19 (P4/16-medium)
+
+   [-1, 1, Conv, [512, 3, 2]],
+   [[-1, 9], 1, Concat, [1]],  # cat head P5
+   [-1, 3, C3, [1024, False]],  # 22 (P5/32-large)
+
+   [[16, 19, 22], 1, Detect, [nc, anchors]],  # Detect(P3, P4, P5)
+  ]

models/yolov5s6.yaml

@@ -0,0 +1,60 @@
+# parameters
+nc: 1  # number of classes
+depth_multiple: 0.33  # model depth multiple
+width_multiple: 0.50  # layer channel multiple
+
+# anchors
+anchors:
+  - [6,7,  9,11,  13,16]  # P3/8
+  - [18,23,  26,33,  37,47]  # P4/16
+  - [54,67,  77,104,  112,154]  # P5/32
+  - [174,238,  258,355,  445,568]  # P6/64
+
+# YOLOv5 backbone
+backbone:
+  # [from, number, module, args]
+  [ [ -1, 1, StemBlock, [ 64, 3, 2 ] ],  # 0-P1/2
+    [ -1, 3, C3, [ 128 ] ],
+    [ -1, 1, Conv, [ 256, 3, 2 ] ],  # 2-P3/8
+    [ -1, 9, C3, [ 256 ] ],
+    [ -1, 1, Conv, [ 512, 3, 2 ] ],  # 4-P4/16
+    [ -1, 9, C3, [ 512 ] ],
+    [ -1, 1, Conv, [ 768, 3, 2 ] ],  # 6-P5/32
+    [ -1, 3, C3, [ 768 ] ],
+    [ -1, 1, Conv, [ 1024, 3, 2 ] ],  # 8-P6/64
+    [ -1, 1, SPP, [ 1024, [ 3, 5, 7 ] ] ],
+    [ -1, 3, C3, [ 1024, False ] ],  # 10
+  ]
+
+# YOLOv5 head
+head:
+  [ [ -1, 1, Conv, [ 768, 1, 1 ] ],
+    [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
+    [ [ -1, 7 ], 1, Concat, [ 1 ] ],  # cat backbone P5
+    [ -1, 3, C3, [ 768, False ] ],  # 14
+
+    [ -1, 1, Conv, [ 512, 1, 1 ] ],
+    [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
+    [ [ -1, 5 ], 1, Concat, [ 1 ] ],  # cat backbone P4
+    [ -1, 3, C3, [ 512, False ] ],  # 18
+
+    [ -1, 1, Conv, [ 256, 1, 1 ] ],
+    [ -1, 1, nn.Upsample, [ None, 2, 'nearest' ] ],
+    [ [ -1, 3 ], 1, Concat, [ 1 ] ],  # cat backbone P3
+    [ -1, 3, C3, [ 256, False ] ],  # 22 (P3/8-small)
+
+    [ -1, 1, Conv, [ 256, 3, 2 ] ],
+    [ [ -1, 19 ], 1, Concat, [ 1 ] ],  # cat head P4
+    [ -1, 3, C3, [ 512, False ] ],  # 25 (P4/16-medium)
+
+    [ -1, 1, Conv, [ 512, 3, 2 ] ],
+    [ [ -1, 15 ], 1, Concat, [ 1 ] ],  # cat head P5
+    [ -1, 3, C3, [ 768, False ] ],  # 28 (P5/32-large)
+
+    [ -1, 1, Conv, [ 768, 3, 2 ] ],
+    [ [ -1, 11 ], 1, Concat, [ 1 ] ],  # cat head P6
+    [ -1, 3, C3, [ 1024, False ] ],  # 31 (P6/64-xlarge)
+
+    [ [ 22, 25, 28, 31 ], 1, Detect, [ nc, anchors ] ],  # Detect(P3, P4, P5, P6)
+  ]
+

onnx_infer.py

@@ -0,0 +1,255 @@
+import onnxruntime
+import numpy as np
+import cv2
+import copy
+import os
+import argparse
+from PIL import Image, ImageDraw, ImageFont
+import time
+
+plateName=r"#京沪津渝冀晋蒙辽吉黑苏浙皖闽赣鲁豫鄂湘粤桂琼川贵云藏陕甘青宁新学警港澳挂使领民航危0123456789ABCDEFGHJKLMNPQRSTUVWXYZ险品"
+mean_value,std_value=((0.588,0.193))#识别模型均值标准差
+
+def decodePlate(preds):        #识别后处理
+    pre=0
+    newPreds=[]
+    for i in range(len(preds)):
+        if preds[i]!=0 and preds[i]!=pre:
+            newPreds.append(preds[i])
+        pre=preds[i]
+    plate=""
+    for i in newPreds:
+        plate+=plateName[int(i)]
+    return plate
+    # return newPreds
+
+def rec_pre_precessing(img,size=(48,168)): #识别前处理
+    img =cv2.resize(img,(168,48))
+    img = img.astype(np.float32)
+    img = (img/255-mean_value)/std_value  #归一化 减均值 除标准差
+    img = img.transpose(2,0,1)         #h,w,c 转为 c,h,w
+    img = img.reshape(1,*img.shape)    #channel,height,width转为batch,channel,height,channel
+    return img
+
+def get_plate_result(img,session_rec): #识别后处理
+    img =rec_pre_precessing(img)
+    y_onnx = session_rec.run([session_rec.get_outputs()[0].name], {session_rec.get_inputs()[0].name: img})[0]
+    # print(y_onnx[0])
+    index =np.argmax(y_onnx[0],axis=1)  #找出概率最大的那个字符的序号
+    # print(y_onnx[0])
+    plate_no = decodePlate(index)
+    # plate_no = decodePlate(y_onnx[0])
+    return plate_no
+
+
+def allFilePath(rootPath,allFIleList):  #遍历文件
+    fileList = os.listdir(rootPath)
+    for temp in fileList:
+        if os.path.isfile(os.path.join(rootPath,temp)):
+            allFIleList.append(os.path.join(rootPath,temp))
+        else:
+            allFilePath(os.path.join(rootPath,temp),allFIleList)
+
+def get_split_merge(img):  #双层车牌进行分割后识别
+    h,w,c = img.shape
+    img_upper = img[0:int(5/12*h),:]
+    img_lower = img[int(1/3*h):,:]
+    img_upper = cv2.resize(img_upper,(img_lower.shape[1],img_lower.shape[0]))
+    new_img = np.hstack((img_upper,img_lower))
+    return new_img
+
+
+def order_points(pts):     # 关键点排列 按照（左上，右上，右下，左下）的顺序排列
+    rect = np.zeros((4, 2), dtype = "float32")
+    s = pts.sum(axis = 1)
+    rect[0] = pts[np.argmin(s)]
+    rect[2] = pts[np.argmax(s)]
+    diff = np.diff(pts, axis = 1)
+    rect[1] = pts[np.argmin(diff)]
+    rect[3] = pts[np.argmax(diff)]
+    return rect
+
+
+def four_point_transform(image, pts):  #透视变换得到矫正后的图像，方便识别
+    rect = order_points(pts)
+    (tl, tr, br, bl) = rect
+    widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
+    widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
+    maxWidth = max(int(widthA), int(widthB))
+    heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
+    heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
+    maxHeight = max(int(heightA), int(heightB))
+    dst = np.array([
+        [0, 0],
+        [maxWidth - 1, 0],
+        [maxWidth - 1, maxHeight - 1],
+        [0, maxHeight - 1]], dtype = "float32")
+    M = cv2.getPerspectiveTransform(rect, dst)
+    warped = cv2.warpPerspective(image, M, (maxWidth, maxHeight))
+ 
+    # return the warped image
+    return warped
+
+def my_letter_box(img,size=(640,640)):  #
+    h,w,c = img.shape
+    r = min(size[0]/h,size[1]/w)
+    new_h,new_w = int(h*r),int(w*r)
+    top = int((size[0]-new_h)/2)
+    left = int((size[1]-new_w)/2)
+    
+    bottom = size[0]-new_h-top
+    right = size[1]-new_w-left
+    img_resize = cv2.resize(img,(new_w,new_h))
+    img = cv2.copyMakeBorder(img_resize,top,bottom,left,right,borderType=cv2.BORDER_CONSTANT,value=(114,114,114))
+    return img,r,left,top
+
+def xywh2xyxy(boxes):   #xywh坐标变为 左上 ，右下坐标 x1,y1  x2,y2
+    xywh =copy.deepcopy(boxes)
+    xywh[:,0]=boxes[:,0]-boxes[:,2]/2
+    xywh[:,1]=boxes[:,1]-boxes[:,3]/2
+    xywh[:,2]=boxes[:,0]+boxes[:,2]/2
+    xywh[:,3]=boxes[:,1]+boxes[:,3]/2
+    return xywh
+ 
+def my_nms(boxes,iou_thresh):         #nms
+    index = np.argsort(boxes[:,4])[::-1]
+    keep = []
+    while index.size >0:
+        i = index[0]
+        keep.append(i)
+        x1=np.maximum(boxes[i,0],boxes[index[1:],0])
+        y1=np.maximum(boxes[i,1],boxes[index[1:],1])
+        x2=np.minimum(boxes[i,2],boxes[index[1:],2])
+        y2=np.minimum(boxes[i,3],boxes[index[1:],3])
+        
+        w = np.maximum(0,x2-x1)
+        h = np.maximum(0,y2-y1)
+
+        inter_area = w*h
+        union_area = (boxes[i,2]-boxes[i,0])*(boxes[i,3]-boxes[i,1])+(boxes[index[1:],2]-boxes[index[1:],0])*(boxes[index[1:],3]-boxes[index[1:],1])
+        iou = inter_area/(union_area-inter_area)
+        idx = np.where(iou<=iou_thresh)[0]
+        index = index[idx+1]
+    return keep
+
+def restore_box(boxes,r,left,top):  #返回原图上面的坐标
+    boxes[:,[0,2,5,7,9,11]]-=left
+    boxes[:,[1,3,6,8,10,12]]-=top
+
+    boxes[:,[0,2,5,7,9,11]]/=r
+    boxes[:,[1,3,6,8,10,12]]/=r
+    return boxes
+
+def detect_pre_precessing(img,img_size):  #检测前处理
+    img,r,left,top=my_letter_box(img,img_size)
+    # cv2.imwrite("1.jpg",img)
+    img =img[:,:,::-1].transpose(2,0,1).copy().astype(np.float32)
+    img=img/255
+    img=img.reshape(1,*img.shape)
+    return img,r,left,top
+
+def post_precessing(dets,r,left,top,conf_thresh=0.3,iou_thresh=0.5):#检测后处理
+    choice = dets[:,:,4]>conf_thresh
+    dets=dets[choice]
+    dets[:,13:15]*=dets[:,4:5]
+    box = dets[:,:4]
+    boxes = xywh2xyxy(box)
+    score= np.max(dets[:,13:15],axis=-1,keepdims=True)
+    index = np.argmax(dets[:,13:15],axis=-1).reshape(-1,1)
+    output = np.concatenate((boxes,score,dets[:,5:13],index),axis=1) 
+    reserve_=my_nms(output,iou_thresh) 
+    output=output[reserve_] 
+    output = restore_box(output,r,left,top)
+    return output
+
+def rec_plate(outputs,img0,session_rec):  #识别车牌
+    dict_list=[]
+    for output in outputs:
+        result_dict={}
+        rect=output[:4].tolist()
+        land_marks = output[5:13].reshape(4,2)
+        roi_img = four_point_transform(img0,land_marks)
+        label = int(output[-1])
+        score = output[4]
+        if label==1:  #代表是双层车牌
+            roi_img = get_split_merge(roi_img)
+        plate_no = get_plate_result(roi_img,session_rec)
+        result_dict['rect']=rect
+        result_dict['landmarks']=land_marks.tolist()
+        result_dict['plate_no']=plate_no
+        result_dict['roi_height']=roi_img.shape[0]
+        dict_list.append(result_dict)
+    return dict_list
+
+def cv2ImgAddText(img, text, left, top, textColor=(0, 255, 0), textSize=20):  #将识别结果画在图上
+    if (isinstance(img, np.ndarray)):  #判断是否OpenCV图片类型
+        img = Image.fromarray(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
+    draw = ImageDraw.Draw(img)
+    fontText = ImageFont.truetype(
+        "fonts/platech.ttf", textSize, encoding="utf-8")
+    draw.text((left, top), text, textColor, font=fontText)
+    return cv2.cvtColor(np.asarray(img), cv2.COLOR_RGB2BGR)
+
+def draw_result(orgimg,dict_list):
+    result_str =""
+    for result in dict_list:
+        rect_area = result['rect']
+        
+        x,y,w,h = rect_area[0],rect_area[1],rect_area[2]-rect_area[0],rect_area[3]-rect_area[1]
+        padding_w = 0.05*w
+        padding_h = 0.11*h
+        rect_area[0]=max(0,int(x-padding_w))
+        rect_area[1]=min(orgimg.shape[1],int(y-padding_h))
+        rect_area[2]=max(0,int(rect_area[2]+padding_w))
+        rect_area[3]=min(orgimg.shape[0],int(rect_area[3]+padding_h))
+
+        height_area = result['roi_height']
+        landmarks=result['landmarks']
+        result = result['plate_no']
+        result_str+=result+" "
+        for i in range(4):  #关键点
+            cv2.circle(orgimg, (int(landmarks[i][0]), int(landmarks[i][1])), 5, clors[i], -1)
+        cv2.rectangle(orgimg,(rect_area[0],rect_area[1]),(rect_area[2],rect_area[3]),(0,0,255),2) #画框
+        if len(result)>=1:
+            orgimg=cv2ImgAddText(orgimg,result,rect_area[0]-height_area,rect_area[1]-height_area-10,(255,0,0),height_area)
+    print(result_str)
+    return orgimg
+
+if __name__ == "__main__":
+    begin = time.time()
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--detect_model',type=str, default=r'weights/plate_detect.onnx', help='model.pt path(s)')  #检测模型
+    parser.add_argument('--rec_model', type=str, default='weights/plate_rec.onnx', help='model.pt path(s)')#识别模型
+    parser.add_argument('--image_path', type=str, default='imgs', help='source') 
+    parser.add_argument('--img_size', type=int, default=640, help='inference size (pixels)')
+    parser.add_argument('--output', type=str, default='result1', help='source') 
+    opt = parser.parse_args()
+    file_list = []
+    allFilePath(opt.image_path,file_list)
+    providers =  ['CPUExecutionProvider']
+    clors = [(255,0,0),(0,255,0),(0,0,255),(255,255,0),(0,255,255)]
+    img_size = (opt.img_size,opt.img_size)
+    session_detect = onnxruntime.InferenceSession(opt.detect_model, providers=providers )
+    session_rec = onnxruntime.InferenceSession(opt.rec_model, providers=providers )
+    if not os.path.exists(opt.output):
+        os.mkdir(opt.output)
+    save_path = opt.output
+    count = 0
+    for pic_ in file_list:
+        count+=1
+        print(count,pic_,end=" ")
+        img=cv2.imread(pic_)
+        img0 = copy.deepcopy(img)
+        img,r,left,top = detect_pre_precessing(img,img_size) #检测前处理
+        # print(img.shape)
+        y_onnx = session_detect.run([session_detect.get_outputs()[0].name], {session_detect.get_inputs()[0].name: img})[0]
+        outputs = post_precessing(y_onnx,r,left,top) #检测后处理
+        result_list=rec_plate(outputs,img0,session_rec)
+        ori_img = draw_result(img0,result_list)
+        img_name = os.path.basename(pic_)
+        save_img_path = os.path.join(save_path,img_name)
+        cv2.imwrite(save_img_path,ori_img)
+    print(f"总共耗时{time.time()-begin} s")
+    
+
+        

openvino_infer.py

@@ -0,0 +1,342 @@
+import cv2
+import matplotlib.pyplot as plt
+import numpy as np
+from openvino.runtime import Core
+import os
+import time
+import copy
+from PIL import Image, ImageDraw, ImageFont
+import argparse
+
+def cv_imread(path):
+    img=cv2.imdecode(np.fromfile(path,dtype=np.uint8),-1)
+    return img
+
+def allFilePath(rootPath,allFIleList):
+    fileList = os.listdir(rootPath)
+    for temp in fileList:
+        if os.path.isfile(os.path.join(rootPath,temp)):
+            # if temp.endswith("jpg"):
+            allFIleList.append(os.path.join(rootPath,temp))
+        else:
+            allFilePath(os.path.join(rootPath,temp),allFIleList)
+
+mean_value,std_value=((0.588,0.193))#识别模型均值标准差
+plateName=r"#京沪津渝冀晋蒙辽吉黑苏浙皖闽赣鲁豫鄂湘粤桂琼川贵云藏陕甘青宁新学警港澳挂使领民航危0123456789ABCDEFGHJKLMNPQRSTUVWXYZ险品"
+
+def rec_pre_precessing(img,size=(48,168)): #识别前处理
+    img =cv2.resize(img,(168,48))
+    img = img.astype(np.float32)
+    img = (img/255-mean_value)/std_value
+    img = img.transpose(2,0,1)
+    img = img.reshape(1,*img.shape)
+    return img
+
+def decodePlate(preds):        #识别后处理
+    pre=0
+    newPreds=[]
+    preds=preds.astype(np.int8)[0]
+    for i in range(len(preds)):
+        if preds[i]!=0 and preds[i]!=pre:
+            newPreds.append(preds[i])
+        pre=preds[i]
+    plate=""
+    for i in newPreds:
+        plate+=plateName[int(i)]
+    return plate
+
+def load_model(onnx_path):
+    ie = Core()
+    model_onnx = ie.read_model(model=onnx_path)
+    compiled_model_onnx = ie.compile_model(model=model_onnx, device_name="CPU")
+    output_layer_onnx = compiled_model_onnx.output(0)
+    return compiled_model_onnx,output_layer_onnx
+
+def get_plate_result(img,rec_model,rec_output):
+    img =rec_pre_precessing(img)
+    # time_b = time.time()
+    res_onnx = rec_model([img])[rec_output]
+    # time_e= time.time()
+    index =np.argmax(res_onnx,axis=-1)  #找出最大概率的那个字符的序号
+    plate_no = decodePlate(index)
+    # print(f'{plate_no},time is {time_e-time_b}')
+    return plate_no
+
+
+def get_split_merge(img):  #双层车牌进行分割后识别
+    h,w,c = img.shape
+    img_upper = img[0:int(5/12*h),:]
+    img_lower = img[int(1/3*h):,:]
+    img_upper = cv2.resize(img_upper,(img_lower.shape[1],img_lower.shape[0]))
+    new_img = np.hstack((img_upper,img_lower))
+    return new_img
+
+
+def order_points(pts):
+    rect = np.zeros((4, 2), dtype = "float32")
+    s = pts.sum(axis = 1)
+    rect[0] = pts[np.argmin(s)]
+    rect[2] = pts[np.argmax(s)]
+    diff = np.diff(pts, axis = 1)
+    rect[1] = pts[np.argmin(diff)]
+    rect[3] = pts[np.argmax(diff)]
+    return rect
+
+
+def four_point_transform(image, pts):
+    rect = order_points(pts)
+    (tl, tr, br, bl) = rect
+    widthA = np.sqrt(((br[0] - bl[0]) ** 2) + ((br[1] - bl[1]) ** 2))
+    widthB = np.sqrt(((tr[0] - tl[0]) ** 2) + ((tr[1] - tl[1]) ** 2))
+    maxWidth = max(int(widthA), int(widthB))
+    heightA = np.sqrt(((tr[0] - br[0]) ** 2) + ((tr[1] - br[1]) ** 2))
+    heightB = np.sqrt(((tl[0] - bl[0]) ** 2) + ((tl[1] - bl[1]) ** 2))
+    maxHeight = max(int(heightA), int(heightB))
+    dst = np.array([
+        [0, 0],
+        [maxWidth - 1, 0],
+        [maxWidth - 1, maxHeight - 1],
+        [0, maxHeight - 1]], dtype = "float32")
+    M = cv2.getPerspectiveTransform(rect, dst)
+    warped = cv2.warpPerspective(image, M, (maxWidth, maxHeight))
+ 
+    # return the warped image
+    return warped
+
+def my_letter_box(img,size=(640,640)):
+    h,w,c = img.shape
+    r = min(size[0]/h,size[1]/w)
+    new_h,new_w = int(h*r),int(w*r)
+    top = int((size[0]-new_h)/2)
+    left = int((size[1]-new_w)/2)
+    
+    bottom = size[0]-new_h-top
+    right = size[1]-new_w-left
+    img_resize = cv2.resize(img,(new_w,new_h))
+    img = cv2.copyMakeBorder(img_resize,top,bottom,left,right,borderType=cv2.BORDER_CONSTANT,value=(114,114,114))
+    return img,r,left,top
+
+def xywh2xyxy(boxes):
+    xywh =copy.deepcopy(boxes)
+    xywh[:,0]=boxes[:,0]-boxes[:,2]/2
+    xywh[:,1]=boxes[:,1]-boxes[:,3]/2
+    xywh[:,2]=boxes[:,0]+boxes[:,2]/2
+    xywh[:,3]=boxes[:,1]+boxes[:,3]/2
+    return xywh
+
+def my_nms(boxes,iou_thresh):
+    index = np.argsort(boxes[:,4])[::-1]
+    keep = []
+    while index.size >0:
+        i = index[0]
+        keep.append(i)
+        x1=np.maximum(boxes[i,0],boxes[index[1:],0])
+        y1=np.maximum(boxes[i,1],boxes[index[1:],1])
+        x2=np.minimum(boxes[i,2],boxes[index[1:],2])
+        y2=np.minimum(boxes[i,3],boxes[index[1:],3])
+        
+        w = np.maximum(0,x2-x1)
+        h = np.maximum(0,y2-y1)
+
+        inter_area = w*h
+        union_area = (boxes[i,2]-boxes[i,0])*(boxes[i,3]-boxes[i,1])+(boxes[index[1:],2]-boxes[index[1:],0])*(boxes[index[1:],3]-boxes[index[1:],1])
+        iou = inter_area/(union_area-inter_area)
+        idx = np.where(iou<=iou_thresh)[0]
+        index = index[idx+1]
+    return keep
+
+def restore_box(boxes,r,left,top):
+    boxes[:,[0,2,5,7,9,11]]-=left
+    boxes[:,[1,3,6,8,10,12]]-=top
+
+    boxes[:,[0,2,5,7,9,11]]/=r
+    boxes[:,[1,3,6,8,10,12]]/=r
+    return boxes
+
+def detect_pre_precessing(img,img_size):
+    img,r,left,top=my_letter_box(img,img_size)
+    # cv2.imwrite("1.jpg",img)
+    img =img[:,:,::-1].transpose(2,0,1).copy().astype(np.float32)
+    img=img/255
+    img=img.reshape(1,*img.shape)
+    return img,r,left,top
+
+def post_precessing(dets,r,left,top,conf_thresh=0.3,iou_thresh=0.5):#检测后处理
+    choice = dets[:,:,4]>conf_thresh
+    dets=dets[choice]
+    dets[:,13:15]*=dets[:,4:5]
+    box = dets[:,:4]
+    boxes = xywh2xyxy(box)
+    score= np.max(dets[:,13:15],axis=-1,keepdims=True)
+    index = np.argmax(dets[:,13:15],axis=-1).reshape(-1,1)
+    output = np.concatenate((boxes,score,dets[:,5:13],index),axis=1) 
+    reserve_=my_nms(output,iou_thresh) 
+    output=output[reserve_] 
+    output = restore_box(output,r,left,top)
+    return output
+
+def rec_plate(outputs,img0,rec_model,rec_output):
+    dict_list=[]
+    for output in outputs:
+        result_dict={}
+        rect=output[:4].tolist()
+        land_marks = output[5:13].reshape(4,2)
+        roi_img = four_point_transform(img0,land_marks)
+        label = int(output[-1])
+        if label==1:  #代表是双层车牌
+            roi_img = get_split_merge(roi_img)
+        plate_no = get_plate_result(roi_img,rec_model,rec_output) #得到车牌识别结果
+        result_dict['rect']=rect
+        result_dict['landmarks']=land_marks.tolist()
+        result_dict['plate_no']=plate_no
+        result_dict['roi_height']=roi_img.shape[0]
+        dict_list.append(result_dict)
+    return dict_list
+
+
+
+def cv2ImgAddText(img, text, left, top, textColor=(0, 255, 0), textSize=20):
+    if (isinstance(img, np.ndarray)):  #判断是否OpenCV图片类型
+        img = Image.fromarray(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
+    draw = ImageDraw.Draw(img)
+    fontText = ImageFont.truetype(
+        "fonts/platech.ttf", textSize, encoding="utf-8")
+    draw.text((left, top), text, textColor, font=fontText)
+    return cv2.cvtColor(np.asarray(img), cv2.COLOR_RGB2BGR)
+
+def draw_result(orgimg,dict_list):
+    result_str =""
+    for result in dict_list:
+        rect_area = result['rect']
+        
+        x,y,w,h = rect_area[0],rect_area[1],rect_area[2]-rect_area[0],rect_area[3]-rect_area[1]
+        padding_w = 0.05*w
+        padding_h = 0.11*h
+        rect_area[0]=max(0,int(x-padding_w))
+        rect_area[1]=min(orgimg.shape[1],int(y-padding_h))
+        rect_area[2]=max(0,int(rect_area[2]+padding_w))
+        rect_area[3]=min(orgimg.shape[0],int(rect_area[3]+padding_h))
+
+        height_area = result['roi_height']
+        landmarks=result['landmarks']
+        result = result['plate_no']
+        result_str+=result+" "
+        # for i in range(4):  #关键点
+        #     cv2.circle(orgimg, (int(landmarks[i][0]), int(landmarks[i][1])), 5, clors[i], -1)
+        
+        if len(result)>=6:
+            cv2.rectangle(orgimg,(rect_area[0],rect_area[1]),(rect_area[2],rect_area[3]),(0,0,255),2) #画框
+            orgimg=cv2ImgAddText(orgimg,result,rect_area[0]-height_area,rect_area[1]-height_area-10,(0,255,0),height_area)
+    # print(result_str)
+    return orgimg
+
+def get_second(capture):
+    if capture.isOpened():
+        rate = capture.get(5)   # 帧速率
+        FrameNumber = capture.get(7)  # 视频文件的帧数
+        duration = FrameNumber/rate  # 帧速率/视频总帧数 是时间，除以60之后单位是分钟
+        return int(rate),int(FrameNumber),int(duration)    
+
+if __name__=="__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--detect_model',type=str, default=r'weights/plate_detect.onnx', help='model.pt path(s)')  #检测模型
+    parser.add_argument('--rec_model', type=str, default='weights/plate_rec.onnx', help='model.pt path(s)')#识别模型
+    parser.add_argument('--image_path', type=str, default='imgs', help='source') 
+    parser.add_argument('--img_size', type=int, default=640, help='inference size (pixels)')
+    parser.add_argument('--output', type=str, default='result1', help='source') 
+    opt = parser.parse_args()
+    file_list=[]
+    file_folder=opt.image_path
+    allFilePath(file_folder,file_list)
+    rec_onnx_path =opt.rec_model
+    detect_onnx_path=opt.detect_model
+    rec_model,rec_output=load_model(rec_onnx_path)
+    detect_model,detect_output=load_model(detect_onnx_path)
+    count=0
+    img_size=(opt.img_size,opt.img_size)
+    begin=time.time()
+    save_path=opt.output
+    if not os.path.exists(save_path):
+        os.mkdir(save_path)
+    for pic_ in file_list:
+    
+        count+=1
+        print(count,pic_,end=" ")
+        img=cv2.imread(pic_)
+        time_b = time.time()
+        if img.shape[-1]==4:
+            img = cv2.cvtColor(img,cv2.COLOR_BGRA2BGR)
+        img0 = copy.deepcopy(img)
+        img,r,left,top = detect_pre_precessing(img,img_size) #检测前处理
+        # print(img.shape)
+        det_result = detect_model([img])[detect_output]
+        outputs = post_precessing(det_result,r,left,top) #检测后处理
+        time_1 = time.time()
+        result_list=rec_plate(outputs,img0,rec_model,rec_output)
+        time_e= time.time()
+        print(f'耗时 {time_e-time_b} s')
+        ori_img = draw_result(img0,result_list)
+        img_name = os.path.basename(pic_)
+        save_img_path = os.path.join(save_path,img_name)
+        
+        cv2.imwrite(save_img_path,ori_img)
+print(f"总共耗时{time.time()-begin} s")
+
+    # video_name = r"plate.mp4"
+    # capture=cv2.VideoCapture(video_name)
+    # fourcc = cv2.VideoWriter_fourcc(*'MP4V') 
+    # fps = capture.get(cv2.CAP_PROP_FPS)  # 帧数
+    # width, height = int(capture.get(cv2.CAP_PROP_FRAME_WIDTH)), int(capture.get(cv2.CAP_PROP_FRAME_HEIGHT))  # 宽高
+    # out = cv2.VideoWriter('2result.mp4', fourcc, fps, (width, height))  # 写入视频
+    # frame_count = 0
+    # fps_all=0
+    # rate,FrameNumber,duration=get_second(capture)
+    # # with open("example.csv",mode='w',newline='') as example_file:
+    #     # fieldnames = ['车牌', '时间']
+    #     # writer = csv.DictWriter(example_file, fieldnames=fieldnames, delimiter=',', quotechar='"', quoting=csv.QUOTE_MINIMAL)
+    #     # writer.writeheader()
+    # if capture.isOpened():
+    #     while True:
+    #         t1 = cv2.getTickCount()
+    #         frame_count+=1
+    #         ret,img=capture.read()
+    #         if not ret:
+    #             break
+    #         # if frame_count%rate==0:
+    #         img0 = copy.deepcopy(img)
+    #         img,r,left,top = detect_pre_precessing(img,img_size) #检测前处理
+    #         # print(img.shape)
+    #         det_result = detect_model([img])[detect_output]
+    #         outputs = post_precessing(det_result,r,left,top) #检测后处理
+    #         result_list=rec_plate(outputs,img0,rec_model,rec_output)
+    #         ori_img = draw_result(img0,result_list)
+    #         t2 =cv2.getTickCount()
+    #         infer_time =(t2-t1)/cv2.getTickFrequency()
+    #         fps=1.0/infer_time
+    #         fps_all+=fps
+    #         str_fps = f'fps:{fps:.4f}'
+    #         out.write(ori_img)
+    #         cv2.putText(ori_img,str_fps,(20,20),cv2.FONT_HERSHEY_SIMPLEX,1,(0,255,0),2)
+    #         cv2.imshow("haha",ori_img)
+    #         cv2.waitKey(1)
+
+    #         # current_time = int(frame_count/FrameNumber*duration)
+    #         # sec = current_time%60
+    #         # minute = current_time//60
+    #         # for result_ in result_list:
+    #         #     plate_no = result_['plate_no']
+    #         #     if not is_car_number(pattern_str,plate_no):
+    #         #         continue
+    #         #     print(f'车牌号:{plate_no},时间:{minute}分{sec}秒')
+    #         #     time_str =f'{minute}分{sec}秒'
+    #         #     writer.writerow({"车牌":plate_no,"时间":time_str})
+    #         # out.write(ori_img)
+            
+            
+    # else:
+    #     print("失败")
+    # capture.release()
+    # out.release()
+    # cv2.destroyAllWindows()
+    # print(f"all frame is {frame_count},average fps is {fps_all/frame_count}")
+

plate_recognition/color_rec.py

@@ -0,0 +1,74 @@
+import warnings
+import cv2
+import torch
+import numpy as np
+import torch.nn as nn
+from torchvision import transforms
+from plate_recognition.plateNet import MyNet_color
+
+
+class MyNet(nn.Module):
+    def __init__(self, class_num=6):
+        super(MyNet, self).__init__()
+        self.class_num = class_num
+        self.backbone = nn.Sequential(
+            nn.Conv2d(in_channels=3, out_channels=16, kernel_size=(5, 5), stride=(1, 1)),  # 0
+            torch.nn.BatchNorm2d(16),
+            nn.ReLU(),
+            nn.MaxPool2d(kernel_size=(2, 2)),
+            nn.Dropout(0),
+            nn.Flatten(),
+            nn.Linear(480, 64),
+            nn.Dropout(0),
+            nn.ReLU(),
+            nn.Linear(64, class_num),
+            nn.Dropout(0),
+            nn.Softmax(1)
+        )
+
+    def forward(self, x):
+        logits = self.backbone(x)
+
+        return logits
+
+
+def init_color_model(model_path,device):
+
+    # device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+    # print("color_rec_device:", device)
+    # PATH = 'E:\study\plate\Chinese_license_plate_detection_recognition-main\weights\color_classify.pth'  # 定义模型路径
+    class_num = 6
+    warnings.filterwarnings('ignore')
+    net = MyNet_color(class_num)
+    net.load_state_dict(torch.load(model_path, map_location=torch.device(device)))
+    net.eval().to(device)
+    modelc = net
+    
+    return modelc
+
+
+def plate_color_rec(img,model,device):
+    class_name = ['黑色', '蓝色', '', '绿色', '白色', '黄色']
+    data_input = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
+    image = cv2.resize(data_input, (34, 9))
+    image = np.transpose(image, (2, 0, 1))
+    img = image / 255
+    img = torch.tensor(img)
+
+    normalize = transforms.Normalize(mean=[0.4243, 0.4947, 0.434],
+                                     std=[0.2569, 0.2478, 0.2174])
+    img = normalize(img)
+    img = torch.unsqueeze(img, dim=0).to(device).float()
+    xx = model(img)
+    
+    return class_name[int(torch.argmax(xx, dim=1)[0])]
+
+
+if __name__ == '__main__':
+    class_name = ['black', 'blue', 'danger', 'green', 'white', 'yellow']
+    data_input = cv2.imread("/mnt/Gpan/Mydata/pytorchPorject/myCrnnPlate/images/test.jpg")  # (高，宽，通道(B，G，R)),（H,W,C）
+    device = torch.device("cuda" if torch.cuda.is_available else "cpu")
+    model = init_color_model("/mnt/Gpan/Mydata/pytorchPorject/Chinese_license_plate_detection_recognition/weights/color_classify.pth",device)
+    color_code = plate_color_rec(data_input,model,device)
+    print(color_code)
+    print(class_name[color_code])

plate_recognition/double_plate_split_merge.py

@@ -0,0 +1,15 @@
+import os
+import cv2
+import numpy as np
+def get_split_merge(img):
+    h,w,c = img.shape
+    img_upper = img[0:int(5/12*h),:]
+    img_lower = img[int(1/3*h):,:]
+    img_upper = cv2.resize(img_upper,(img_lower.shape[1],img_lower.shape[0]))
+    new_img = np.hstack((img_upper,img_lower))
+    return new_img
+
+if __name__=="__main__":
+    img = cv2.imread("double_plate/tmp8078.png")
+    new_img =get_split_merge(img)
+    cv2.imwrite("double_plate/new.jpg",new_img)

plate_recognition/plateNet.py

@@ -0,0 +1,128 @@
+import torch.nn as nn
+import torch
+
+
+class myNet_ocr(nn.Module):
+    def __init__(self,cfg=None,num_classes=78,export=False):
+        super(myNet_ocr, self).__init__()
+        if cfg is None:
+            cfg =[32,32,64,64,'M',128,128,'M',196,196,'M',256,256]
+            # cfg =[32,32,'M',64,64,'M',128,128,'M',256,256]
+        self.feature = self.make_layers(cfg, True)
+        self.export = export
+        # self.classifier = nn.Linear(cfg[-1], num_classes)
+        # self.loc =  nn.MaxPool2d((2, 2), (5, 1), (0, 1),ceil_mode=True)
+        # self.loc =  nn.AvgPool2d((2, 2), (5, 2), (0, 1),ceil_mode=False)
+        self.loc =  nn.MaxPool2d((5, 2), (1, 1),(0,1),ceil_mode=False)
+        self.newCnn=nn.Conv2d(cfg[-1],num_classes,1,1)
+        # self.newBn=nn.BatchNorm2d(num_classes)
+    def make_layers(self, cfg, batch_norm=False):
+        layers = []
+        in_channels = 3
+        for i in range(len(cfg)):
+            if i == 0:
+                conv2d =nn.Conv2d(in_channels, cfg[i], kernel_size=5,stride =1)
+                if batch_norm:
+                    layers += [conv2d, nn.BatchNorm2d(cfg[i]), nn.ReLU(inplace=True)]
+                else:
+                    layers += [conv2d, nn.ReLU(inplace=True)]
+                in_channels = cfg[i]
+            else :
+                if cfg[i] == 'M':
+                    layers += [nn.MaxPool2d(kernel_size=3, stride=2,ceil_mode=True)]
+                else:
+                    conv2d = nn.Conv2d(in_channels, cfg[i], kernel_size=3, padding=(1,1),stride =1)
+                    if batch_norm:
+                        layers += [conv2d, nn.BatchNorm2d(cfg[i]), nn.ReLU(inplace=True)]
+                    else:
+                        layers += [conv2d, nn.ReLU(inplace=True)]
+                    in_channels = cfg[i]
+        return nn.Sequential(*layers)
+
+    def forward(self, x):
+        x = self.feature(x)
+        x=self.loc(x)
+        x=self.newCnn(x)
+        # x=self.newBn(x)
+        if self.export:
+            conv = x.squeeze(2) # b *512 * width
+            conv = conv.transpose(2,1)  # [w, b, c]
+            conv =conv.argmax(dim=2)
+            return conv
+        else:
+            b, c, h, w = x.size()
+            assert h == 1, "the height of conv must be 1"
+            conv = x.squeeze(2) # b *512 * width
+            conv = conv.permute(2, 0, 1)  # [w, b, c]
+            # output = F.log_softmax(self.rnn(conv), dim=2)
+            output = torch.softmax(conv, dim=2)
+            return output
+
+myCfg = [32,'M',64,'M',96,'M',128,'M',256]
+class myNet(nn.Module):
+    def __init__(self,cfg=None,num_classes=3):
+        super(myNet, self).__init__()
+        if cfg is None:
+            cfg = myCfg
+        self.feature = self.make_layers(cfg, True)
+        self.classifier = nn.Linear(cfg[-1], num_classes)
+    def make_layers(self, cfg, batch_norm=False):
+        layers = []
+        in_channels = 3
+        for i in range(len(cfg)):
+            if i == 0:
+                conv2d =nn.Conv2d(in_channels, cfg[i], kernel_size=5,stride =1)
+                if batch_norm:
+                    layers += [conv2d, nn.BatchNorm2d(cfg[i]), nn.ReLU(inplace=True)]
+                else:
+                    layers += [conv2d, nn.ReLU(inplace=True)]
+                in_channels = cfg[i]
+            else :
+                if cfg[i] == 'M':
+                    layers += [nn.MaxPool2d(kernel_size=3, stride=2,ceil_mode=True)]
+                else:
+                    conv2d = nn.Conv2d(in_channels, cfg[i], kernel_size=3, padding=1,stride =1)
+                    if batch_norm:
+                        layers += [conv2d, nn.BatchNorm2d(cfg[i]), nn.ReLU(inplace=True)]
+                    else:
+                        layers += [conv2d, nn.ReLU(inplace=True)]
+                    in_channels = cfg[i]
+        return nn.Sequential(*layers)
+
+    def forward(self, x):
+        x = self.feature(x)
+        x = nn.AvgPool2d(kernel_size=3, stride=1)(x)
+        x = x.view(x.size(0), -1)
+        y = self.classifier(x)
+        return y
+    
+    
+class MyNet_color(nn.Module):
+    def __init__(self, class_num=6):
+        super(MyNet_color, self).__init__()
+        self.class_num = class_num
+        self.backbone = nn.Sequential(
+            nn.Conv2d(in_channels=3, out_channels=16, kernel_size=(5, 5), stride=(1, 1)),  # 0
+            torch.nn.BatchNorm2d(16),
+            nn.ReLU(),
+            nn.MaxPool2d(kernel_size=(2, 2)),
+            nn.Dropout(0),
+            nn.Flatten(),
+            nn.Linear(480, 64),
+            nn.Dropout(0),
+            nn.ReLU(),
+            nn.Linear(64, class_num),
+            nn.Dropout(0),
+            nn.Softmax(1)
+        )
+
+    def forward(self, x):
+        logits = self.backbone(x)
+
+        return logits
+
+if __name__ == '__main__':
+    x = torch.randn(1,3,48,216)
+    model = myNet_ocr(num_classes=78,export=True)
+    out = model(x)
+    print(out.shape)

plate_recognition/plate_rec.py

@@ -0,0 +1,99 @@
+from plate_recognition.plateNet import myNet_ocr
+import torch
+import torch.nn as nn
+import cv2
+import numpy as np
+import os
+import time
+import sys
+
+def cv_imread(path):  #可以读取中文路径的图片
+    img=cv2.imdecode(np.fromfile(path,dtype=np.uint8),-1)
+    return img
+
+def allFilePath(rootPath,allFIleList):
+    fileList = os.listdir(rootPath)
+    for temp in fileList:
+        if os.path.isfile(os.path.join(rootPath,temp)):
+            if temp.endswith('.jpg') or temp.endswith('.png') or temp.endswith('.JPG'):
+                allFIleList.append(os.path.join(rootPath,temp))
+        else:
+            allFilePath(os.path.join(rootPath,temp),allFIleList)
+device = torch.device('cuda') if torch.cuda.is_available() else torch.device("cpu")
+plateName=r"#京沪津渝冀晋蒙辽吉黑苏浙皖闽赣鲁豫鄂湘粤桂琼川贵云藏陕甘青宁新学警港澳挂使领民航危0123456789ABCDEFGHJKLMNPQRSTUVWXYZ险品"
+mean_value,std_value=(0.588,0.193)
+def decodePlate(preds):
+    pre=0
+    newPreds=[]
+    for i in range(len(preds)):
+        if preds[i]!=0 and preds[i]!=pre:
+            newPreds.append(preds[i])
+        pre=preds[i]
+    return newPreds
+
+def image_processing(img,device):
+    img = cv2.resize(img, (168,48))
+    img = np.reshape(img, (48, 168, 3))
+
+    # normalize
+    img = img.astype(np.float32)
+    img = (img / 255. - mean_value) / std_value
+    img = img.transpose([2, 0, 1])
+    img = torch.from_numpy(img)
+
+    img = img.to(device)
+    img = img.view(1, *img.size())
+    return img
+
+def get_plate_result(img,device,model):
+    input = image_processing(img,device)
+    preds = model(input)
+    # preds =preds.argmax(dim=2) #找出概率最大的那个字符
+    # print(preds)
+    preds=preds.view(-1).detach().cpu().numpy()
+    newPreds=decodePlate(preds)
+    plate=""
+    for i in newPreds:
+        plate+=plateName[i]
+    # if not (plate[0] in plateName[1:44] ):
+    #     return ""
+    return plate
+
+def init_model(device,model_path):
+    # print( print(sys.path))
+    # model_path ="plate_recognition/model/checkpoint_61_acc_0.9715.pth"
+    check_point = torch.load(model_path,map_location=device)
+    model_state=check_point['state_dict']
+    cfg=check_point['cfg']
+    model_path = os.sep.join([sys.path[0],model_path])
+    model = myNet_ocr(num_classes=len(plateName),export=True,cfg=cfg)
+   
+    model.load_state_dict(model_state)
+    model.to(device)
+    model.eval()
+    return model
+
+# model = init_model(device)
+if __name__ == '__main__':
+
+   image_path ="images/tmp2424.png"
+   testPath = r"double_plate"
+   fileList=[]
+   allFilePath(testPath,fileList)
+#    result = get_plate_result(image_path,device)
+#    print(result)
+   model = init_model(device)
+   right=0
+   begin = time.time()
+   for imge_path in fileList:
+        plate=get_plate_result(imge_path)
+        plate_ori = imge_path.split('/')[-1].split('_')[0]
+        # print(plate,"---",plate_ori)
+        if(plate==plate_ori):
+
+            right+=1
+        else:
+            print(plate_ori,"--->",plate,imge_path)
+   end=time.time()
+   print("sum:%d ,right:%d , accuracy: %f, time: %f"%(len(fileList),right,right/len(fileList),end-begin))
+        

requirements.txt

@@ -0,0 +1,47 @@
+asttokens
+backcall
+charset-normalizer
+cycler
+dataclasses
+debugpy
+decorator
+executing
+fonttools
+idna
+ipykernel
+ipython
+jedi
+jupyter-client
+jupyter-core
+kiwisolver
+matplotlib
+matplotlib-inline
+nest-asyncio
+numpy
+opencv-python
+packaging
+pandas
+parso
+pickleshare
+Pillow
+prompt-toolkit
+psutil
+pure-eval
+Pygments
+pyparsing
+python-dateutil
+pytz
+PyYAML
+pyzmq
+requests
+scipy
+seaborn
+six
+stack-data
+thop
+tornado
+tqdm
+traitlets
+typing-extensions
+urllib3
+wcwidth

test.py

@@ -0,0 +1,336 @@
+import argparse
+import json
+import os
+from pathlib import Path
+from threading import Thread
+
+import numpy as np
+import torch
+import yaml
+from tqdm import tqdm
+
+from models.experimental import attempt_load
+from utils.datasets import create_dataloader
+from utils.general import coco80_to_coco91_class, check_dataset, check_file, check_img_size, box_iou, \
+    non_max_suppression, scale_coords, xyxy2xywh, xywh2xyxy, set_logging, increment_path, non_max_suppression_face
+from utils.loss import compute_loss
+from utils.metrics import ap_per_class, ConfusionMatrix
+from utils.plots import plot_images, output_to_target, plot_study_txt
+from utils.torch_utils import select_device, time_synchronized
+
+
+def test(data,
+         weights=None,
+         batch_size=32,
+         imgsz=640,
+         conf_thres=0.001,
+         iou_thres=0.6,  # for NMS
+         save_json=False,
+         single_cls=False,
+         augment=False,
+         verbose=False,
+         model=None,
+         dataloader=None,
+         save_dir=Path(''),  # for saving images
+         save_txt=False,  # for auto-labelling
+         save_hybrid=False,  # for hybrid auto-labelling
+         save_conf=False,  # save auto-label confidences
+         plots=True,
+         log_imgs=0):  # number of logged images
+
+    # Initialize/load model and set device
+    training = model is not None
+    if training:  # called by train.py
+        device = next(model.parameters()).device  # get model device
+
+    else:  # called directly
+        set_logging()
+        device = select_device(opt.device, batch_size=batch_size)
+
+        # Directories
+        save_dir = Path(increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok))  # increment run
+        (save_dir / 'labels' if save_txt else save_dir).mkdir(parents=True, exist_ok=True)  # make dir
+
+        # Load model
+        model = attempt_load(weights, map_location=device)  # load FP32 model
+        imgsz = check_img_size(imgsz, s=model.stride.max())  # check img_size
+
+        # Multi-GPU disabled, incompatible with .half() https://github.com/ultralytics/yolov5/issues/99
+        # if device.type != 'cpu' and torch.cuda.device_count() > 1:
+        #     model = nn.DataParallel(model)
+
+    # Half
+    half = device.type != 'cpu'  # half precision only supported on CUDA
+    if half:
+        model.half()
+
+    # Configure
+    model.eval()
+    is_coco = data.endswith('coco.yaml')  # is COCO dataset
+    with open(data) as f:
+        data = yaml.load(f, Loader=yaml.FullLoader)  # model dict
+    check_dataset(data)  # check
+    nc = 1 if single_cls else int(data['nc'])  # number of classes
+    iouv = torch.linspace(0.5, 0.95, 10).to(device)  # iou vector for mAP@0.5:0.95
+    niou = iouv.numel()
+
+    # Logging
+    log_imgs, wandb = min(log_imgs, 100), None  # ceil
+    try:
+        import wandb  # Weights & Biases
+    except ImportError:
+        log_imgs = 0
+
+    # Dataloader
+    if not training:
+        img = torch.zeros((1, 3, imgsz, imgsz), device=device)  # init img
+        _ = model(img.half() if half else img) if device.type != 'cpu' else None  # run once
+        path = data['test'] if opt.task == 'test' else data['val']  # path to val/test images
+        dataloader = create_dataloader(path, imgsz, batch_size, model.stride.max(), opt, pad=0.5, rect=True)[0]
+
+    seen = 0
+    confusion_matrix = ConfusionMatrix(nc=nc)
+    names = {k: v for k, v in enumerate(model.names if hasattr(model, 'names') else model.module.names)}
+    coco91class = coco80_to_coco91_class()
+    s = ('%20s' + '%12s' * 6) % ('Class', 'Images', 'Targets', 'P', 'R', 'mAP@.5', 'mAP@.5:.95')
+    p, r, f1, mp, mr, map50, map, t0, t1 = 0., 0., 0., 0., 0., 0., 0., 0., 0.
+    loss = torch.zeros(3, device=device)
+    jdict, stats, ap, ap_class, wandb_images = [], [], [], [], []
+    for batch_i, (img, targets, paths, shapes) in enumerate(tqdm(dataloader, desc=s)):
+        img = img.to(device, non_blocking=True)
+        img = img.half() if half else img.float()  # uint8 to fp16/32
+        img /= 255.0  # 0 - 255 to 0.0 - 1.0
+        targets = targets.to(device)
+        nb, _, height, width = img.shape  # batch size, channels, height, width
+
+        with torch.no_grad():
+            # Run model
+            t = time_synchronized()
+            inf_out, train_out = model(img, augment=augment)  # inference and training outputs
+            t0 += time_synchronized() - t
+
+            # Compute loss
+            if training:
+                loss += compute_loss([x.float() for x in train_out], targets, model)[1][:3]  # box, obj, cls
+
+            # Run NMS
+            targets[:, 2:6] *= torch.Tensor([width, height, width, height]).to(device)  # to pixels
+            lb = [targets[targets[:, 0] == i, 1:] for i in range(nb)] if save_hybrid else []  # for autolabelling
+            t = time_synchronized()
+            #output = non_max_suppression(inf_out, conf_thres=conf_thres, iou_thres=iou_thres, labels=lb)
+            output = non_max_suppression_face(inf_out, conf_thres=conf_thres, iou_thres=iou_thres, labels=lb)
+            t1 += time_synchronized() - t
+
+        # Statistics per image
+        for si, pred in enumerate(output):
+            pred = torch.cat((pred[:, :5], pred[:, 13:]), 1) # throw landmark in thresh
+            labels = targets[targets[:, 0] == si, 1:]
+            nl = len(labels)
+            tcls = labels[:, 0].tolist() if nl else []  # target class
+            path = Path(paths[si])
+            seen += 1
+
+            if len(pred) == 0:
+                if nl:
+                    stats.append((torch.zeros(0, niou, dtype=torch.bool), torch.Tensor(), torch.Tensor(), tcls))
+                continue
+
+            # Predictions
+            predn = pred.clone()
+            scale_coords(img[si].shape[1:], predn[:, :4], shapes[si][0], shapes[si][1])  # native-space pred
+
+            # Append to text file
+            if save_txt:
+                gn = torch.tensor(shapes[si][0])[[1, 0, 1, 0]]  # normalization gain whwh
+                for *xyxy, conf, cls in predn.tolist():
+                    xywh = (xyxy2xywh(torch.tensor(xyxy).view(1, 4)) / gn).view(-1).tolist()  # normalized xywh
+                    line = (cls, *xywh, conf) if save_conf else (cls, *xywh)  # label format
+                    with open(save_dir / 'labels' / (path.stem + '.txt'), 'a') as f:
+                        f.write(('%g ' * len(line)).rstrip() % line + '\n')
+
+            # W&B logging
+            if plots and len(wandb_images) < log_imgs:
+                box_data = [{"position": {"minX": xyxy[0], "minY": xyxy[1], "maxX": xyxy[2], "maxY": xyxy[3]},
+                             "class_id": int(cls),
+                             "box_caption": "%s %.3f" % (names[cls], conf),
+                             "scores": {"class_score": conf},
+                             "domain": "pixel"} for *xyxy, conf, cls in pred.tolist()]
+                boxes = {"predictions": {"box_data": box_data, "class_labels": names}}  # inference-space
+                wandb_images.append(wandb.Image(img[si], boxes=boxes, caption=path.name))
+
+            # Append to pycocotools JSON dictionary
+            if save_json:
+                # [{"image_id": 42, "category_id": 18, "bbox": [258.15, 41.29, 348.26, 243.78], "score": 0.236}, ...
+                image_id = int(path.stem) if path.stem.isnumeric() else path.stem
+                box = xyxy2xywh(predn[:, :4])  # xywh
+                box[:, :2] -= box[:, 2:] / 2  # xy center to top-left corner
+                for p, b in zip(pred.tolist(), box.tolist()):
+                    jdict.append({'image_id': image_id,
+                                  'category_id': coco91class[int(p[15])] if is_coco else int(p[15]),
+                                  'bbox': [round(x, 3) for x in b],
+                                  'score': round(p[4], 5)})
+
+            # Assign all predictions as incorrect
+            correct = torch.zeros(pred.shape[0], niou, dtype=torch.bool, device=device)
+            if nl:
+                detected = []  # target indices
+                tcls_tensor = labels[:, 0]
+
+                # target boxes
+                tbox = xywh2xyxy(labels[:, 1:5])
+                scale_coords(img[si].shape[1:], tbox, shapes[si][0], shapes[si][1])  # native-space labels
+                if plots:
+                    confusion_matrix.process_batch(pred, torch.cat((labels[:, 0:1], tbox), 1))
+
+                # Per target class
+                for cls in torch.unique(tcls_tensor):
+                    ti = (cls == tcls_tensor).nonzero(as_tuple=False).view(-1)  # prediction indices
+                    pi = (cls == pred[:, 5]).nonzero(as_tuple=False).view(-1)  # target indices
+
+                    # Search for detections
+                    if pi.shape[0]:
+                        # Prediction to target ious
+                        ious, i = box_iou(predn[pi, :4], tbox[ti]).max(1)  # best ious, indices
+
+                        # Append detections
+                        detected_set = set()
+                        for j in (ious > iouv[0]).nonzero(as_tuple=False):
+                            d = ti[i[j]]  # detected target
+                            if d.item() not in detected_set:
+                                detected_set.add(d.item())
+                                detected.append(d)
+                                correct[pi[j]] = ious[j] > iouv  # iou_thres is 1xn
+                                if len(detected) == nl:  # all targets already located in image
+                                    break
+
+            # Append statistics (correct, conf, pcls, tcls)
+            stats.append((correct.cpu(), pred[:, 4].cpu(), pred[:, 5].cpu(), tcls))
+
+        # Plot images
+        if plots and batch_i < 3:
+            f = save_dir / f'test_batch{batch_i}_labels.jpg'  # labels
+            Thread(target=plot_images, args=(img, targets, paths, f, names), daemon=True).start()
+            f = save_dir / f'test_batch{batch_i}_pred.jpg'  # predictions
+            Thread(target=plot_images, args=(img, output_to_target(output), paths, f, names), daemon=True).start()
+
+    # Compute statistics
+    stats = [np.concatenate(x, 0) for x in zip(*stats)]  # to numpy
+    if len(stats) and stats[0].any():
+        p, r, ap, f1, ap_class = ap_per_class(*stats, plot=plots, save_dir=save_dir, names=names)
+        p, r, ap50, ap = p[:, 0], r[:, 0], ap[:, 0], ap.mean(1)  # [P, R, AP@0.5, AP@0.5:0.95]
+        mp, mr, map50, map = p.mean(), r.mean(), ap50.mean(), ap.mean()
+        nt = np.bincount(stats[3].astype(np.int64), minlength=nc)  # number of targets per class
+    else:
+        nt = torch.zeros(1)
+
+    # Print results
+    pf = '%20s' + '%12.3g' * 6  # print format
+    print(pf % ('all', seen, nt.sum(), mp, mr, map50, map))
+
+    # Print results per class
+    if verbose and nc > 1 and len(stats):
+        for i, c in enumerate(ap_class):
+            print(pf % (names[c], seen, nt[c], p[i], r[i], ap50[i], ap[i]))
+
+    # Print speeds
+    t = tuple(x / seen * 1E3 for x in (t0, t1, t0 + t1)) + (imgsz, imgsz, batch_size)  # tuple
+    if not training:
+        print('Speed: %.1f/%.1f/%.1f ms inference/NMS/total per %gx%g image at batch-size %g' % t)
+
+    # Plots
+    if plots:
+        confusion_matrix.plot(save_dir=save_dir, names=list(names.values()))
+        if wandb and wandb.run:
+            wandb.log({"Images": wandb_images})
+            wandb.log({"Validation": [wandb.Image(str(f), caption=f.name) for f in sorted(save_dir.glob('test*.jpg'))]})
+
+    # Save JSON
+    if save_json and len(jdict):
+        w = Path(weights[0] if isinstance(weights, list) else weights).stem if weights is not None else ''  # weights
+        anno_json = '../coco/annotations/instances_val2017.json'  # annotations json
+        pred_json = str(save_dir / f"{w}_predictions.json")  # predictions json
+        print('\nEvaluating pycocotools mAP... saving %s...' % pred_json)
+        with open(pred_json, 'w') as f:
+            json.dump(jdict, f)
+
+        try:  # https://github.com/cocodataset/cocoapi/blob/master/PythonAPI/pycocoEvalDemo.ipynb
+            from pycocotools.coco import COCO
+            from pycocotools.cocoeval import COCOeval
+
+            anno = COCO(anno_json)  # init annotations api
+            pred = anno.loadRes(pred_json)  # init predictions api
+            eval = COCOeval(anno, pred, 'bbox')
+            if is_coco:
+                eval.params.imgIds = [int(Path(x).stem) for x in dataloader.dataset.img_files]  # image IDs to evaluate
+            eval.evaluate()
+            eval.accumulate()
+            eval.summarize()
+            map, map50 = eval.stats[:2]  # update results (mAP@0.5:0.95, mAP@0.5)
+        except Exception as e:
+            print(f'pycocotools unable to run: {e}')
+
+    # Return results
+    if not training:
+        s = f"\n{len(list(save_dir.glob('labels/*.txt')))} labels saved to {save_dir / 'labels'}" if save_txt else ''
+        print(f"Results saved to {save_dir}{s}")
+    model.float()  # for training
+    maps = np.zeros(nc) + map
+    for i, c in enumerate(ap_class):
+        maps[c] = ap[i]
+    return (mp, mr, map50, map, *(loss.cpu() / len(dataloader)).tolist()), maps, t
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser(prog='test.py')
+    parser.add_argument('--weights', nargs='+', type=str, default='yolov5s.pt', help='model.pt path(s)')
+    parser.add_argument('--data', type=str, default='data/coco128.yaml', help='*.data path')
+    parser.add_argument('--batch-size', type=int, default=32, help='size of each image batch')
+    parser.add_argument('--img-size', type=int, default=640, help='inference size (pixels)')
+    parser.add_argument('--conf-thres', type=float, default=0.001, help='object confidence threshold')
+    parser.add_argument('--iou-thres', type=float, default=0.6, help='IOU threshold for NMS')
+    parser.add_argument('--task', default='val', help="'val', 'test', 'study'")
+    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--single-cls', action='store_true', help='treat as single-class dataset')
+    parser.add_argument('--augment', action='store_true', help='augmented inference')
+    parser.add_argument('--verbose', action='store_true', help='report mAP by class')
+    parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')
+    parser.add_argument('--save-hybrid', action='store_true', help='save label+prediction hybrid results to *.txt')
+    parser.add_argument('--save-conf', action='store_true', help='save confidences in --save-txt labels')
+    parser.add_argument('--save-json', action='store_true', help='save a cocoapi-compatible JSON results file')
+    parser.add_argument('--project', default='runs/test', help='save to project/name')
+    parser.add_argument('--name', default='exp', help='save to project/name')
+    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
+    opt = parser.parse_args()
+    opt.save_json |= opt.data.endswith('coco.yaml')
+    opt.data = check_file(opt.data)  # check file
+    print(opt)
+
+    if opt.task in ['val', 'test']:  # run normally
+        test(opt.data,
+             opt.weights,
+             opt.batch_size,
+             opt.img_size,
+             opt.conf_thres,
+             opt.iou_thres,
+             opt.save_json,
+             opt.single_cls,
+             opt.augment,
+             opt.verbose,
+             save_txt=opt.save_txt | opt.save_hybrid,
+             save_hybrid=opt.save_hybrid,
+             save_conf=opt.save_conf,
+             )
+
+    elif opt.task == 'study':  # run over a range of settings and save/plot
+        for weights in ['yolov5s.pt', 'yolov5m.pt', 'yolov5l.pt', 'yolov5x.pt']:
+            f = 'study_%s_%s.txt' % (Path(opt.data).stem, Path(weights).stem)  # filename to save to
+            x = list(range(320, 800, 64))  # x axis
+            y = []  # y axis
+            for i in x:  # img-size
+                print('\nRunning %s point %s...' % (f, i))
+                r, _, t = test(opt.data, weights, opt.batch_size, i, opt.conf_thres, opt.iou_thres, opt.save_json,
+                               plots=False)
+                y.append(r + t)  # results and times
+            np.savetxt(f, y, fmt='%10.4g')  # save
+        os.system('zip -r study.zip study_*.txt')
+        plot_study_txt(f, x)  # plot

test_widerface.py

@@ -0,0 +1,170 @@
+import argparse
+import glob
+import time
+from pathlib import Path
+
+import os
+import cv2
+import torch
+import torch.backends.cudnn as cudnn
+from numpy import random
+import numpy as np
+from models.experimental import attempt_load
+from utils.datasets import letterbox
+from utils.general import check_img_size, check_requirements, non_max_suppression_face, apply_classifier, \
+    scale_coords, xyxy2xywh, strip_optimizer, set_logging, increment_path
+from utils.plots import plot_one_box
+from utils.torch_utils import select_device, load_classifier, time_synchronized
+from tqdm import tqdm
+
+def dynamic_resize(shape, stride=64):
+    max_size = max(shape[0], shape[1])
+    if max_size % stride != 0:
+        max_size = (int(max_size / stride) + 1) * stride 
+    return max_size
+
+def scale_coords_landmarks(img1_shape, coords, img0_shape, ratio_pad=None):
+    # Rescale coords (xyxy) from img1_shape to img0_shape
+    if ratio_pad is None:  # calculate from img0_shape
+        gain = min(img1_shape[0] / img0_shape[0], img1_shape[1] / img0_shape[1])  # gain  = old / new
+        pad = (img1_shape[1] - img0_shape[1] * gain) / 2, (img1_shape[0] - img0_shape[0] * gain) / 2  # wh padding
+    else:
+        gain = ratio_pad[0][0]
+        pad = ratio_pad[1]
+
+    coords[:, [0, 2, 4, 6, 8]] -= pad[0]  # x padding
+    coords[:, [1, 3, 5, 7, 9]] -= pad[1]  # y padding
+    coords[:, :10] /= gain
+    #clip_coords(coords, img0_shape)
+    coords[:, 0].clamp_(0, img0_shape[1])  # x1
+    coords[:, 1].clamp_(0, img0_shape[0])  # y1
+    coords[:, 2].clamp_(0, img0_shape[1])  # x2
+    coords[:, 3].clamp_(0, img0_shape[0])  # y2
+    coords[:, 4].clamp_(0, img0_shape[1])  # x3
+    coords[:, 5].clamp_(0, img0_shape[0])  # y3
+    coords[:, 6].clamp_(0, img0_shape[1])  # x4
+    coords[:, 7].clamp_(0, img0_shape[0])  # y4
+    coords[:, 8].clamp_(0, img0_shape[1])  # x5
+    coords[:, 9].clamp_(0, img0_shape[0])  # y5
+    return coords
+
+def show_results(img, xywh, conf, landmarks, class_num):
+    h,w,c = img.shape
+    tl = 1 or round(0.002 * (h + w) / 2) + 1  # line/font thickness
+    x1 = int(xywh[0] * w - 0.5 * xywh[2] * w)
+    y1 = int(xywh[1] * h - 0.5 * xywh[3] * h)
+    x2 = int(xywh[0] * w + 0.5 * xywh[2] * w)
+    y2 = int(xywh[1] * h + 0.5 * xywh[3] * h)
+    cv2.rectangle(img, (x1,y1), (x2, y2), (0,255,0), thickness=tl, lineType=cv2.LINE_AA)
+
+    clors = [(255,0,0),(0,255,0),(0,0,255),(255,255,0),(0,255,255)]
+
+    for i in range(5):
+        point_x = int(landmarks[2 * i] * w)
+        point_y = int(landmarks[2 * i + 1] * h)
+        cv2.circle(img, (point_x, point_y), tl+1, clors[i], -1)
+
+    tf = max(tl - 1, 1)  # font thickness
+    label = str(int(class_num)) + ': ' + str(conf)[:5]
+    cv2.putText(img, label, (x1, y1 - 2), 0, tl / 3, [225, 255, 255], thickness=tf, lineType=cv2.LINE_AA)
+    return img
+
+def detect(model, img0):
+    stride = int(model.stride.max())  # model stride
+    imgsz = opt.img_size
+    if imgsz <= 0:                    # original size    
+        imgsz = dynamic_resize(img0.shape)
+    imgsz = check_img_size(imgsz, s=64)  # check img_size
+    img = letterbox(img0, imgsz)[0]
+    # Convert
+    img = img[:, :, ::-1].transpose(2, 0, 1)  # BGR to RGB, to 3x416x416
+    img = np.ascontiguousarray(img)
+    img = torch.from_numpy(img).to(device)
+    img = img.float()  # uint8 to fp16/32
+    img /= 255.0  # 0 - 255 to 0.0 - 1.0
+    if img.ndimension() == 3:
+        img = img.unsqueeze(0)
+
+    # Inference
+    pred = model(img, augment=opt.augment)[0]
+    # Apply NMS
+    pred = non_max_suppression_face(pred, opt.conf_thres, opt.iou_thres)[0]
+    gn = torch.tensor(img0.shape)[[1, 0, 1, 0]].to(device)  # normalization gain whwh
+    gn_lks = torch.tensor(img0.shape)[[1, 0, 1, 0, 1, 0, 1, 0, 1, 0]].to(device)  # normalization gain landmarks
+    boxes = []
+    h, w, c = img0.shape
+    if pred is not None:
+        pred[:, :4] = scale_coords(img.shape[2:], pred[:, :4], img0.shape).round()
+        pred[:, 5:15] = scale_coords_landmarks(img.shape[2:], pred[:, 5:15], img0.shape).round()
+        for j in range(pred.size()[0]):
+            xywh = (xyxy2xywh(pred[j, :4].view(1, 4)) / gn).view(-1)
+            xywh = xywh.data.cpu().numpy()
+            conf = pred[j, 4].cpu().numpy()
+            landmarks = (pred[j, 5:15].view(1, 10) / gn_lks).view(-1).tolist()
+            class_num = pred[j, 15].cpu().numpy()
+            x1 = int(xywh[0] * w - 0.5 * xywh[2] * w)
+            y1 = int(xywh[1] * h - 0.5 * xywh[3] * h)
+            x2 = int(xywh[0] * w + 0.5 * xywh[2] * w)
+            y2 = int(xywh[1] * h + 0.5 * xywh[3] * h)
+            boxes.append([x1, y1, x2-x1, y2-y1, conf])
+    return boxes
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--weights', nargs='+', type=str, default='runs/train/exp5/weights/last.pt', help='model.pt path(s)')
+    parser.add_argument('--img-size', type=int, default=640, help='inference size (pixels)')
+    parser.add_argument('--conf-thres', type=float, default=0.02, help='object confidence threshold')
+    parser.add_argument('--iou-thres', type=float, default=0.5, help='IOU threshold for NMS')
+    parser.add_argument('--device', default='0', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS')
+    parser.add_argument('--augment', action='store_true', help='augmented inference')
+    parser.add_argument('--update', action='store_true', help='update all models')
+    parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --class 0, or --class 0 2 3')
+    parser.add_argument('--project', default='runs/detect', help='save results to project/name')
+    parser.add_argument('--name', default='exp', help='save results to project/name')
+    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
+    parser.add_argument('--save_folder', default='./widerface_evaluate/widerface_txt/', type=str, help='Dir to save txt results')
+    parser.add_argument('--dataset_folder', default='../WiderFace/val/images/', type=str, help='dataset path')
+    parser.add_argument('--folder_pict', default='/yolov5-face/data/widerface/val/wider_val.txt', type=str, help='folder_pict')
+    opt = parser.parse_args()
+    print(opt)
+
+    # changhy : read folder_pict
+    pict_folder = {}
+    with open(opt.folder_pict, 'r') as f:
+        lines = f.readlines()
+        for line in lines:
+            line = line.strip().split('/')
+            pict_folder[line[-1]] = line[-2]
+
+    # Load model
+    device = select_device(opt.device)
+    model = attempt_load(opt.weights, map_location=device)  # load FP32 model
+    with torch.no_grad():
+        # testing dataset
+        testset_folder = opt.dataset_folder
+
+        for image_path in tqdm(glob.glob(os.path.join(testset_folder, '*'))):
+            if image_path.endswith('.txt'):
+                continue
+            img0 = cv2.imread(image_path)  # BGR
+            if img0 is None:
+                print(f'ignore : {image_path}')
+                continue
+            boxes = detect(model, img0)
+            # --------------------------------------------------------------------
+            image_name = os.path.basename(image_path)
+            txt_name = os.path.splitext(image_name)[0] + ".txt"
+            save_name = os.path.join(opt.save_folder, pict_folder[image_name], txt_name)
+            dirname = os.path.dirname(save_name)
+            if not os.path.isdir(dirname):
+                os.makedirs(dirname)
+            with open(save_name, "w") as fd:
+                file_name = os.path.basename(save_name)[:-4] + "\n"            
+                bboxs_num = str(len(boxes)) + "\n"
+                fd.write(file_name)
+                fd.write(bboxs_num)
+                for box in boxes:
+                    fd.write('%d %d %d %d %.03f' % (box[0], box[1], box[2], box[3], box[4] if box[4] <= 1 else 1) + '\n')
+        print('done.')

train.py

@@ -0,0 +1,602 @@
+import argparse
+import logging
+import math
+import os
+import random
+import time
+from pathlib import Path
+from threading import Thread
+from warnings import warn
+
+import numpy as np
+import torch.distributed as dist
+import torch.nn as nn
+import torch.nn.functional as F
+import torch.optim as optim
+import torch.optim.lr_scheduler as lr_scheduler
+import torch.utils.data
+import yaml
+from torch.cuda import amp
+from torch.nn.parallel import DistributedDataParallel as DDP
+from torch.utils.tensorboard import SummaryWriter
+from tqdm import tqdm
+
+import test  # import test.py to get mAP after each epoch
+from models.experimental import attempt_load
+from models.yolo import Model
+from utils.autoanchor import check_anchors
+from utils.face_datasets import create_dataloader
+from utils.general import labels_to_class_weights, increment_path, labels_to_image_weights, init_seeds, \
+    fitness, strip_optimizer, get_latest_run, check_dataset, check_file, check_git_status, check_img_size, \
+    print_mutation, set_logging
+from utils.google_utils import attempt_download
+from utils.loss import compute_loss
+from utils.plots import plot_images, plot_labels, plot_results, plot_evolution
+from utils.torch_utils import ModelEMA, select_device, intersect_dicts, torch_distributed_zero_first
+
+logger = logging.getLogger(__name__)
+begin_save=1
+try:
+    import wandb
+except ImportError:
+    wandb = None
+    logger.info("Install Weights & Biases for experiment logging via 'pip install wandb' (recommended)")
+
+
+def train(hyp, opt, device, tb_writer=None, wandb=None):
+    logger.info(f'Hyperparameters {hyp}')
+    save_dir, epochs, batch_size, total_batch_size, weights, rank = \
+        Path(opt.save_dir), opt.epochs, opt.batch_size, opt.total_batch_size, opt.weights, opt.global_rank
+
+    # Directories
+    wdir = save_dir / 'weights'
+    wdir.mkdir(parents=True, exist_ok=True)  # make dir
+    last = wdir / 'last.pt'
+    best = wdir / 'best.pt'
+    results_file = save_dir / 'results.txt'
+
+    # Save run settings
+    with open(save_dir / 'hyp.yaml', 'w') as f:
+        yaml.dump(hyp, f, sort_keys=False)
+    with open(save_dir / 'opt.yaml', 'w') as f:
+        yaml.dump(vars(opt), f, sort_keys=False)
+
+    # Configure
+    plots = not opt.evolve  # create plots
+    cuda = device.type != 'cpu'
+    init_seeds(2 + rank)
+    with open(opt.data) as f:
+        data_dict = yaml.load(f, Loader=yaml.FullLoader)  # data dict
+    with torch_distributed_zero_first(rank):
+        check_dataset(data_dict)  # check
+    train_path = data_dict['train']
+    test_path = data_dict['val']
+    nc = 1 if opt.single_cls else int(data_dict['nc'])  # number of classes
+    names = ['item'] if opt.single_cls and len(data_dict['names']) != 1 else data_dict['names']  # class names
+    assert len(names) == nc, '%g names found for nc=%g dataset in %s' % (len(names), nc, opt.data)  # check
+
+    # Model
+    pretrained = weights.endswith('.pt')
+    if pretrained:
+        with torch_distributed_zero_first(rank):
+            attempt_download(weights)  # download if not found locally
+        ckpt = torch.load(weights, map_location=device)  # load checkpoint
+        if hyp.get('anchors'):
+            ckpt['model'].yaml['anchors'] = round(hyp['anchors'])  # force autoanchor
+        model = Model(opt.cfg or ckpt['model'].yaml, ch=3, nc=nc).to(device)  # create
+        exclude = ['anchor'] if opt.cfg or hyp.get('anchors') else []  # exclude keys
+        state_dict = ckpt['model'].float().state_dict()  # to FP32
+        state_dict = intersect_dicts(state_dict, model.state_dict(), exclude=exclude)  # intersect
+        model.load_state_dict(state_dict, strict=False)  # load
+        logger.info('Transferred %g/%g items from %s' % (len(state_dict), len(model.state_dict()), weights))  # report
+    else:
+        model = Model(opt.cfg, ch=3, nc=nc).to(device)  # create
+
+    # Freeze
+    freeze = []  # parameter names to freeze (full or partial)
+    for k, v in model.named_parameters():
+        v.requires_grad = True  # train all layers
+        if any(x in k for x in freeze):
+            print('freezing %s' % k)
+            v.requires_grad = False
+
+    # Optimizer
+    nbs = 64  # nominal batch size
+    accumulate = max(round(nbs / total_batch_size), 1)  # accumulate loss before optimizing
+    hyp['weight_decay'] *= total_batch_size * accumulate / nbs  # scale weight_decay
+
+    pg0, pg1, pg2 = [], [], []  # optimizer parameter groups
+    for k, v in model.named_modules():
+        if hasattr(v, 'bias') and isinstance(v.bias, nn.Parameter):
+            pg2.append(v.bias)  # biases
+        if isinstance(v, nn.BatchNorm2d):
+            pg0.append(v.weight)  # no decay
+        elif hasattr(v, 'weight') and isinstance(v.weight, nn.Parameter):
+            pg1.append(v.weight)  # apply decay
+
+    if opt.adam:
+        optimizer = optim.Adam(pg0, lr=hyp['lr0'], betas=(hyp['momentum'], 0.999))  # adjust beta1 to momentum
+    else:
+        optimizer = optim.SGD(pg0, lr=hyp['lr0'], momentum=hyp['momentum'], nesterov=True)
+
+    optimizer.add_param_group({'params': pg1, 'weight_decay': hyp['weight_decay']})  # add pg1 with weight_decay
+    optimizer.add_param_group({'params': pg2})  # add pg2 (biases)
+    logger.info('Optimizer groups: %g .bias, %g conv.weight, %g other' % (len(pg2), len(pg1), len(pg0)))
+    del pg0, pg1, pg2
+
+    # Scheduler https://arxiv.org/pdf/1812.01187.pdf
+    # https://pytorch.org/docs/stable/_modules/torch/optim/lr_scheduler.html#OneCycleLR
+    lf = lambda x: ((1 + math.cos(x * math.pi / epochs)) / 2) * (1 - hyp['lrf']) + hyp['lrf']  # cosine
+    scheduler = lr_scheduler.LambdaLR(optimizer, lr_lambda=lf)
+    # plot_lr_scheduler(optimizer, scheduler, epochs)
+
+    # Logging
+    if wandb and wandb.run is None:
+        opt.hyp = hyp  # add hyperparameters
+        wandb_run = wandb.init(config=opt, resume="allow",
+                               project='YOLOv5' if opt.project == 'runs/train' else Path(opt.project).stem,
+                               name=save_dir.stem,
+                               id=ckpt.get('wandb_id') if 'ckpt' in locals() else None)
+    loggers = {'wandb': wandb}  # loggers dict
+
+    # Resume
+    start_epoch, best_fitness = 0, 0.0
+    if pretrained:
+        # Optimizer
+        if ckpt['optimizer'] is not None:
+            optimizer.load_state_dict(ckpt['optimizer'])
+            best_fitness = 0
+
+        # Results
+        if ckpt.get('training_results') is not None:
+            with open(results_file, 'w') as file:
+                file.write(ckpt['training_results'])  # write results.txt
+
+        # Epochs
+        # start_epoch = ckpt['epoch'] + 1
+        if opt.resume:
+            assert start_epoch > 0, '%s training to %g epochs is finished, nothing to resume.' % (weights, epochs)
+        if epochs < start_epoch:
+            logger.info('%s has been trained for %g epochs. Fine-tuning for %g additional epochs.' %
+                        (weights, ckpt['epoch'], epochs))
+            epochs += ckpt['epoch']  # finetune additional epochs
+
+        del ckpt, state_dict
+
+    # Image sizes
+    gs = int(max(model.stride))  # grid size (max stride)
+    imgsz, imgsz_test = [check_img_size(x, gs) for x in opt.img_size]  # verify imgsz are gs-multiples
+
+    # DP mode
+    if cuda and rank == -1 and torch.cuda.device_count() > 1:
+        model = torch.nn.DataParallel(model)
+
+    # SyncBatchNorm
+    if opt.sync_bn and cuda and rank != -1:
+        model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model).to(device)
+        logger.info('Using SyncBatchNorm()')
+
+    # EMA
+    ema = ModelEMA(model) if rank in [-1, 0] else None
+
+    # DDP mode
+    if cuda and rank != -1:
+        model = DDP(model, device_ids=[opt.local_rank], output_device=opt.local_rank)
+
+    # Trainloader
+    dataloader, dataset = create_dataloader(train_path, imgsz, batch_size, gs, opt,
+                                            hyp=hyp, augment=True, cache=opt.cache_images, rect=opt.rect, rank=rank,
+                                            world_size=opt.world_size, workers=opt.workers,
+                                            image_weights=opt.image_weights)
+    mlc = np.concatenate(dataset.labels, 0)[:, 0].max()  # max label class
+    nb = len(dataloader)  # number of batches
+    assert mlc < nc, 'Label class %g exceeds nc=%g in %s. Possible class labels are 0-%g' % (mlc, nc, opt.data, nc - 1)
+
+    # Process 0
+    if rank in [-1, 0]:
+        ema.updates = start_epoch * nb // accumulate  # set EMA updates
+        testloader = create_dataloader(test_path, imgsz_test, total_batch_size, gs, opt,  # testloader
+                                       hyp=hyp, cache=opt.cache_images and not opt.notest, rect=True,
+                                       rank=-1, world_size=opt.world_size, workers=opt.workers, pad=0.5)[0]
+
+        if not opt.resume:
+            labels = np.concatenate(dataset.labels, 0)
+            c = torch.tensor(labels[:, 0])  # classes
+            # cf = torch.bincount(c.long(), minlength=nc) + 1.  # frequency
+            # model._initialize_biases(cf.to(device))
+            if plots:
+                plot_labels(labels, save_dir, loggers)
+                if tb_writer:
+                    tb_writer.add_histogram('classes', c, 0)
+
+            # Anchors
+            if not opt.noautoanchor:
+                check_anchors(dataset, model=model, thr=hyp['anchor_t'], imgsz=imgsz)
+
+    # Model parameters
+    hyp['cls'] *= nc / 80.  # scale coco-tuned hyp['cls'] to current dataset
+    model.nc = nc  # attach number of classes to model
+    model.hyp = hyp  # attach hyperparameters to model
+    model.gr = 1.0  # iou loss ratio (obj_loss = 1.0 or iou)
+    model.class_weights = labels_to_class_weights(dataset.labels, nc).to(device) * nc  # attach class weights
+    model.names = names
+
+    # Start training
+    t0 = time.time()
+    nw = max(round(hyp['warmup_epochs'] * nb), 1000)  # number of warmup iterations, max(3 epochs, 1k iterations)
+    # nw = min(nw, (epochs - start_epoch) / 2 * nb)  # limit warmup to < 1/2 of training
+    maps = np.zeros(nc)  # mAP per class
+    results = (0, 0, 0, 0, 0, 0, 0)  # P, R, mAP@.5, mAP@.5-.95, val_loss(box, obj, cls)
+    scheduler.last_epoch = start_epoch - 1  # do not move
+    scaler = amp.GradScaler(enabled=cuda)
+    logger.info('Image sizes %g train, %g test\n'
+                'Using %g dataloader workers\nLogging results to %s\n'
+                'Starting training for %g epochs...' % (imgsz, imgsz_test, dataloader.num_workers, save_dir, epochs))
+    for epoch in range(start_epoch, epochs):  # epoch ------------------------------------------------------------------
+        model.train()
+
+        # Update image weights (optional)
+        if opt.image_weights:
+            # Generate indices
+            if rank in [-1, 0]:
+                cw = model.class_weights.cpu().numpy() * (1 - maps) ** 2 / nc  # class weights
+                iw = labels_to_image_weights(dataset.labels, nc=nc, class_weights=cw)  # image weights
+                dataset.indices = random.choices(range(dataset.n), weights=iw, k=dataset.n)  # rand weighted idx
+            # Broadcast if DDP
+            if rank != -1:
+                indices = (torch.tensor(dataset.indices) if rank == 0 else torch.zeros(dataset.n)).int()
+                dist.broadcast(indices, 0)
+                if rank != 0:
+                    dataset.indices = indices.cpu().numpy()
+
+        # Update mosaic border
+        # b = int(random.uniform(0.25 * imgsz, 0.75 * imgsz + gs) // gs * gs)
+        # dataset.mosaic_border = [b - imgsz, -b]  # height, width borders
+
+        mloss = torch.zeros(5, device=device)  # mean losses
+        if rank != -1:
+            dataloader.sampler.set_epoch(epoch)
+        pbar = enumerate(dataloader)
+        logger.info(('\n' + '%10s' * 9) % ('Epoch', 'gpu_mem', 'box', 'obj', 'cls', 'landmark', 'total', 'targets', 'img_size'))
+        if rank in [-1, 0]:
+            pbar = tqdm(pbar, total=nb)  # progress bar
+        optimizer.zero_grad()
+        for i, (imgs, targets, paths, _) in pbar:  # batch -------------------------------------------------------------
+            ni = i + nb * epoch  # number integrated batches (since train start)
+            imgs = imgs.to(device, non_blocking=True).float() / 255.0  # uint8 to float32, 0-255 to 0.0-1.0
+
+            # Warmup
+            if ni <= nw:
+                xi = [0, nw]  # x interp
+                # model.gr = np.interp(ni, xi, [0.0, 1.0])  # iou loss ratio (obj_loss = 1.0 or iou)
+                accumulate = max(1, np.interp(ni, xi, [1, nbs / total_batch_size]).round())
+                for j, x in enumerate(optimizer.param_groups):
+                    # bias lr falls from 0.1 to lr0, all other lrs rise from 0.0 to lr0
+                    x['lr'] = np.interp(ni, xi, [hyp['warmup_bias_lr'] if j == 2 else 0.0, x['initial_lr'] * lf(epoch)])
+                    if 'momentum' in x:
+                        x['momentum'] = np.interp(ni, xi, [hyp['warmup_momentum'], hyp['momentum']])
+
+            # Multi-scale
+            if opt.multi_scale:
+                sz = random.randrange(imgsz * 0.5, imgsz * 1.5 + gs) // gs * gs  # size
+                sf = sz / max(imgs.shape[2:])  # scale factor
+                if sf != 1:
+                    ns = [math.ceil(x * sf / gs) * gs for x in imgs.shape[2:]]  # new shape (stretched to gs-multiple)
+                    imgs = F.interpolate(imgs, size=ns, mode='bilinear', align_corners=False)
+
+            # Forward
+            with amp.autocast(enabled=cuda):
+                pred = model(imgs)  # forward
+                loss, loss_items = compute_loss(pred, targets.to(device), model)  # loss scaled by batch_size
+                if rank != -1:
+                    loss *= opt.world_size  # gradient averaged between devices in DDP mode
+
+            # Backward
+            scaler.scale(loss).backward()
+
+            # Optimize
+            if ni % accumulate == 0:
+                scaler.step(optimizer)  # optimizer.step
+                scaler.update()
+                optimizer.zero_grad()
+                if ema:
+                    ema.update(model)
+
+            # Print
+            if rank in [-1, 0]:
+                mloss = (mloss * i + loss_items) / (i + 1)  # update mean losses
+                mem = '%.3gG' % (torch.cuda.memory_reserved() / 1E9 if torch.cuda.is_available() else 0)  # (GB)
+                s = ('%10s' * 2 + '%10.4g' * 7) % (
+                    '%g/%g' % (epoch, epochs - 1), mem, *mloss, targets.shape[0], imgs.shape[-1])
+                pbar.set_description(s)
+
+                # Plot
+                if plots and ni < 3:
+                    f = save_dir / f'train_batch{ni}.jpg'  # filename
+                    Thread(target=plot_images, args=(imgs, targets, paths, f), daemon=True).start()
+                    # if tb_writer:
+                    #     tb_writer.add_image(f, result, dataformats='HWC', global_step=epoch)
+                    #     tb_writer.add_graph(model, imgs)  # add model to tensorboard
+                elif plots and ni == 3 and wandb:
+                    wandb.log({"Mosaics": [wandb.Image(str(x), caption=x.name) for x in save_dir.glob('train*.jpg')]})
+
+            # end batch ------------------------------------------------------------------------------------------------
+        # end epoch ----------------------------------------------------------------------------------------------------
+
+        # Scheduler
+        lr = [x['lr'] for x in optimizer.param_groups]  # for tensorboard
+        scheduler.step()
+
+        # DDP process 0 or single-GPU
+        if rank in [-1, 0] and epoch > begin_save:
+            # mAP
+            if ema:
+                ema.update_attr(model, include=['yaml', 'nc', 'hyp', 'gr', 'names', 'stride', 'class_weights'])
+            final_epoch = epoch + 1 == epochs
+            if not opt.notest or final_epoch:  # Calculate mAP
+                results, maps, times = test.test(opt.data,
+                                                 batch_size=total_batch_size,
+                                                 imgsz=imgsz_test,
+                                                 model=ema.ema,
+                                                 single_cls=opt.single_cls,
+                                                 dataloader=testloader,
+                                                 save_dir=save_dir,
+                                                 plots=False,
+                                                 log_imgs=opt.log_imgs if wandb else 0)
+
+            # Write
+            with open(results_file, 'a') as f:
+                f.write(s + '%10.4g' * 7 % results + '\n')  # P, R, mAP@.5, mAP@.5-.95, val_loss(box, obj, cls)
+            if len(opt.name) and opt.bucket:
+                os.system('gsutil cp %s gs://%s/results/results%s.txt' % (results_file, opt.bucket, opt.name))
+
+            # Log
+            tags = ['train/box_loss', 'train/obj_loss', 'train/cls_loss',  # train loss
+                    'metrics/precision', 'metrics/recall', 'metrics/mAP_0.5', 'metrics/mAP_0.5:0.95',
+                    'val/box_loss', 'val/obj_loss', 'val/cls_loss',  # val loss
+                    'x/lr0', 'x/lr1', 'x/lr2']  # params
+            for x, tag in zip(list(mloss[:-1]) + list(results) + lr, tags):
+                if tb_writer:
+                    tb_writer.add_scalar(tag, x, epoch)  # tensorboard
+                if wandb:
+                    wandb.log({tag: x})  # W&B
+
+            # Update best mAP
+            fi = fitness(np.array(results).reshape(1, -1))  # weighted combination of [P, R, mAP@.5, mAP@.5-.95]
+            if fi > best_fitness:
+                best_fitness = fi
+
+            # Save model
+            save = (not opt.nosave) or (final_epoch and not opt.evolve)
+            if save:
+                with open(results_file, 'r') as f:  # create checkpoint
+                    ckpt = {'epoch': epoch,
+                            'best_fitness': best_fitness,
+                            'training_results': f.read(),
+                            'model': ema.ema,
+                            'optimizer': None if final_epoch else optimizer.state_dict(),
+                            'wandb_id': wandb_run.id if wandb else None}
+
+                # Save last, best and delete
+                torch.save(ckpt, last)
+                if best_fitness == fi:
+                    ckpt_best = {
+                            'epoch': epoch,
+                            'best_fitness': best_fitness,
+                            # 'training_results': f.read(),
+                            'model': ema.ema,
+                            # 'optimizer': None if final_epoch else optimizer.state_dict(),
+                            # 'wandb_id': wandb_run.id if wandb else None
+                            }
+                    torch.save(ckpt_best, best)
+                del ckpt
+        # end epoch ----------------------------------------------------------------------------------------------------
+    # end training
+
+    if rank in [-1, 0]:
+        # Strip optimizers
+        final = best if best.exists() else last  # final model
+        for f in [last, best]:
+            if f.exists():
+                strip_optimizer(f)  # strip optimizers
+        if opt.bucket:
+            os.system(f'gsutil cp {final} gs://{opt.bucket}/weights')  # upload
+
+        # Plots
+        if plots:
+            plot_results(save_dir=save_dir)  # save as results.png
+            if wandb:
+                files = ['results.png', 'precision_recall_curve.png', 'confusion_matrix.png']
+                wandb.log({"Results": [wandb.Image(str(save_dir / f), caption=f) for f in files
+                                       if (save_dir / f).exists()]})
+                if opt.log_artifacts:
+                    wandb.log_artifact(artifact_or_path=str(final), type='model', name=save_dir.stem)
+
+        # Test best.pt
+        logger.info('%g epochs completed in %.3f hours.\n' % (epoch - start_epoch + 1, (time.time() - t0) / 3600))
+        if opt.data.endswith('coco.yaml') and nc == 80:  # if COCO
+            for conf, iou, save_json in ([0.25, 0.45, False], [0.001, 0.65, True]):  # speed, mAP tests
+                results, _, _ = test.test(opt.data,
+                                          batch_size=total_batch_size,
+                                          imgsz=imgsz_test,
+                                          conf_thres=conf,
+                                          iou_thres=iou,
+                                          model=attempt_load(final, device).half(),
+                                          single_cls=opt.single_cls,
+                                          dataloader=testloader,
+                                          save_dir=save_dir,
+                                          save_json=save_json,
+                                          plots=False)
+
+    else:
+        dist.destroy_process_group()
+
+    wandb.run.finish() if wandb and wandb.run else None
+    torch.cuda.empty_cache()
+    return results
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--weights', type=str, default='weights/yolov5s.pt', help='initial weights path')
+    parser.add_argument('--cfg', type=str, default='models/yolov5s.yaml', help='model.yaml path')
+    parser.add_argument('--data', type=str, default='data/widerface.yaml', help='data.yaml path')
+    parser.add_argument('--hyp', type=str, default='data/hyp.scratch.yaml', help='hyperparameters path')
+    parser.add_argument('--epochs', type=int, default=120)
+    parser.add_argument('--batch-size', type=int, default=32, help='total batch size for all GPUs')
+    parser.add_argument('--img-size', nargs='+', type=int, default=[640, 640], help='[train, test] image sizes')
+    parser.add_argument('--rect', action='store_true', help='rectangular training')
+    parser.add_argument('--resume', nargs='?', const=True, default=False, help='resume most recent training')
+    parser.add_argument('--nosave', action='store_true', help='only save final checkpoint')
+    parser.add_argument('--notest', action='store_true', help='only test final epoch')
+    parser.add_argument('--noautoanchor', action='store_true', help='disable autoanchor check')
+    parser.add_argument('--evolve', action='store_true', help='evolve hyperparameters')
+    parser.add_argument('--bucket', type=str, default='', help='gsutil bucket')
+    parser.add_argument('--cache-images', action='store_true', help='cache images for faster training')
+    parser.add_argument('--image-weights', action='store_true', help='use weighted image selection for training')
+    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
+    parser.add_argument('--multi-scale', action='store_true', default=True, help='vary img-size +/- 50%%')
+    parser.add_argument('--single-cls', action='store_true', help='train multi-class data as single-class')
+    parser.add_argument('--adam', action='store_true', help='use torch.optim.Adam() optimizer')
+    parser.add_argument('--sync-bn', action='store_true', help='use SyncBatchNorm, only available in DDP mode')
+    parser.add_argument('--local_rank', type=int, default=-1, help='DDP parameter, do not modify')
+    parser.add_argument('--log-imgs', type=int, default=16, help='number of images for W&B logging, max 100')
+    parser.add_argument('--log-artifacts', action='store_true', help='log artifacts, i.e. final trained model')
+    parser.add_argument('--workers', type=int, default=4, help='maximum number of dataloader workers')
+    parser.add_argument('--project', default='runs/train', help='save to project/name')
+    parser.add_argument('--name', default='exp', help='save to project/name')
+    parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
+    opt = parser.parse_args()
+
+    # Set DDP variables
+    opt.total_batch_size = opt.batch_size
+    opt.world_size = int(os.environ['WORLD_SIZE']) if 'WORLD_SIZE' in os.environ else 1
+    opt.global_rank = int(os.environ['RANK']) if 'RANK' in os.environ else -1
+    set_logging(opt.global_rank)
+    if opt.global_rank in [-1, 0]:
+        check_git_status()
+
+    # Resume
+    if opt.resume:  # resume an interrupted run
+        ckpt = opt.resume if isinstance(opt.resume, str) else get_latest_run()  # specified or most recent path
+        assert os.path.isfile(ckpt), 'ERROR: --resume checkpoint does not exist'
+        with open(Path(ckpt).parent.parent / 'opt.yaml') as f:
+            opt = argparse.Namespace(**yaml.load(f, Loader=yaml.FullLoader))  # replace
+        opt.cfg, opt.weights, opt.resume = '', ckpt, True
+        logger.info('Resuming training from %s' % ckpt)
+    else:
+        # opt.hyp = opt.hyp or ('hyp.finetune.yaml' if opt.weights else 'hyp.scratch.yaml')
+        opt.data, opt.cfg, opt.hyp = check_file(opt.data), check_file(opt.cfg), check_file(opt.hyp)  # check files
+        assert len(opt.cfg) or len(opt.weights), 'either --cfg or --weights must be specified'
+        opt.img_size.extend([opt.img_size[-1]] * (2 - len(opt.img_size)))  # extend to 2 sizes (train, test)
+        opt.name = 'evolve' if opt.evolve else opt.name
+        opt.save_dir = increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok | opt.evolve)  # increment run
+
+    # DDP mode
+    device = select_device(opt.device, batch_size=opt.batch_size)
+    if opt.local_rank != -1:
+        assert torch.cuda.device_count() > opt.local_rank
+        torch.cuda.set_device(opt.local_rank)
+        device = torch.device('cuda', opt.local_rank)
+        dist.init_process_group(backend='nccl', init_method='env://')  # distributed backend
+        assert opt.batch_size % opt.world_size == 0, '--batch-size must be multiple of CUDA device count'
+        opt.batch_size = opt.total_batch_size // opt.world_size
+
+    # Hyperparameters
+    with open(opt.hyp) as f:
+        hyp = yaml.load(f, Loader=yaml.FullLoader)  # load hyps
+        if 'box' not in hyp:
+            warn('Compatibility: %s missing "box" which was renamed from "giou" in %s' %
+                 (opt.hyp, 'https://github.com/ultralytics/yolov5/pull/1120'))
+            hyp['box'] = hyp.pop('giou')
+
+    # Train
+    logger.info(opt)
+    if not opt.evolve:
+        tb_writer = None  # init loggers
+        if opt.global_rank in [-1, 0]:
+            logger.info(f'Start Tensorboard with "tensorboard --logdir {opt.project}", view at http://localhost:6006/')
+            tb_writer = SummaryWriter(opt.save_dir)  # Tensorboard
+        train(hyp, opt, device, tb_writer, wandb)
+
+    # Evolve hyperparameters (optional)
+    else:
+        # Hyperparameter evolution metadata (mutation scale 0-1, lower_limit, upper_limit)
+        meta = {'lr0': (1, 1e-5, 1e-1),  # initial learning rate (SGD=1E-2, Adam=1E-3)
+                'lrf': (1, 0.01, 1.0),  # final OneCycleLR learning rate (lr0 * lrf)
+                'momentum': (0.3, 0.6, 0.98),  # SGD momentum/Adam beta1
+                'weight_decay': (1, 0.0, 0.001),  # optimizer weight decay
+                'warmup_epochs': (1, 0.0, 5.0),  # warmup epochs (fractions ok)
+                'warmup_momentum': (1, 0.0, 0.95),  # warmup initial momentum
+                'warmup_bias_lr': (1, 0.0, 0.2),  # warmup initial bias lr
+                'box': (1, 0.02, 0.2),  # box loss gain
+                'cls': (1, 0.2, 4.0),  # cls loss gain
+                'cls_pw': (1, 0.5, 2.0),  # cls BCELoss positive_weight
+                'obj': (1, 0.2, 4.0),  # obj loss gain (scale with pixels)
+                'obj_pw': (1, 0.5, 2.0),  # obj BCELoss positive_weight
+                'iou_t': (0, 0.1, 0.7),  # IoU training threshold
+                'anchor_t': (1, 2.0, 8.0),  # anchor-multiple threshold
+                'anchors': (2, 2.0, 10.0),  # anchors per output grid (0 to ignore)
+                'fl_gamma': (0, 0.0, 2.0),  # focal loss gamma (efficientDet default gamma=1.5)
+                'hsv_h': (1, 0.0, 0.1),  # image HSV-Hue augmentation (fraction)
+                'hsv_s': (1, 0.0, 0.9),  # image HSV-Saturation augmentation (fraction)
+                'hsv_v': (1, 0.0, 0.9),  # image HSV-Value augmentation (fraction)
+                'degrees': (1, 0.0, 45.0),  # image rotation (+/- deg)
+                'translate': (1, 0.0, 0.9),  # image translation (+/- fraction)
+                'scale': (1, 0.0, 0.9),  # image scale (+/- gain)
+                'shear': (1, 0.0, 10.0),  # image shear (+/- deg)
+                'perspective': (0, 0.0, 0.001),  # image perspective (+/- fraction), range 0-0.001
+                'flipud': (1, 0.0, 1.0),  # image flip up-down (probability)
+                'fliplr': (0, 0.0, 1.0),  # image flip left-right (probability)
+                'mosaic': (1, 0.0, 1.0),  # image mixup (probability)
+                'mixup': (1, 0.0, 1.0)}  # image mixup (probability)
+
+        assert opt.local_rank == -1, 'DDP mode not implemented for --evolve'
+        opt.notest, opt.nosave = True, True  # only test/save final epoch
+        # ei = [isinstance(x, (int, float)) for x in hyp.values()]  # evolvable indices
+        yaml_file = Path(opt.save_dir) / 'hyp_evolved.yaml'  # save best result here
+        if opt.bucket:
+            os.system('gsutil cp gs://%s/evolve.txt .' % opt.bucket)  # download evolve.txt if exists
+
+        for _ in range(300):  # generations to evolve
+            if Path('evolve.txt').exists():  # if evolve.txt exists: select best hyps and mutate
+                # Select parent(s)
+                parent = 'single'  # parent selection method: 'single' or 'weighted'
+                x = np.loadtxt('evolve.txt', ndmin=2)
+                n = min(5, len(x))  # number of previous results to consider
+                x = x[np.argsort(-fitness(x))][:n]  # top n mutations
+                w = fitness(x) - fitness(x).min()  # weights
+                if parent == 'single' or len(x) == 1:
+                    # x = x[random.randint(0, n - 1)]  # random selection
+                    x = x[random.choices(range(n), weights=w)[0]]  # weighted selection
+                elif parent == 'weighted':
+                    x = (x * w.reshape(n, 1)).sum(0) / w.sum()  # weighted combination
+
+                # Mutate
+                mp, s = 0.8, 0.2  # mutation probability, sigma
+                npr = np.random
+                npr.seed(int(time.time()))
+                g = np.array([x[0] for x in meta.values()])  # gains 0-1
+                ng = len(meta)
+                v = np.ones(ng)
+                while all(v == 1):  # mutate until a change occurs (prevent duplicates)
+                    v = (g * (npr.random(ng) < mp) * npr.randn(ng) * npr.random() * s + 1).clip(0.3, 3.0)
+                for i, k in enumerate(hyp.keys()):  # plt.hist(v.ravel(), 300)
+                    hyp[k] = float(x[i + 7] * v[i])  # mutate
+
+            # Constrain to limits
+            for k, v in meta.items():
+                hyp[k] = max(hyp[k], v[1])  # lower limit
+                hyp[k] = min(hyp[k], v[2])  # upper limit
+                hyp[k] = round(hyp[k], 5)  # significant digits
+
+            # Train mutation
+            results = train(hyp.copy(), opt, device, wandb=wandb)
+
+            # Write mutation results
+            print_mutation(hyp.copy(), results, yaml_file, opt.bucket)
+
+        # Plot results
+        plot_evolution(yaml_file)
+        print(f'Hyperparameter evolution complete. Best results saved as: {yaml_file}\n'
+              f'Command to train a new model with these hyperparameters: $ python train.py --hyp {yaml_file}')

utils/activations.py

@@ -0,0 +1,72 @@
+# Activation functions
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+# SiLU https://arxiv.org/pdf/1606.08415.pdf ----------------------------------------------------------------------------
+class SiLU(nn.Module):  # export-friendly version of nn.SiLU()
+    @staticmethod
+    def forward(x):
+        return x * torch.sigmoid(x)
+
+
+class Hardswish(nn.Module):  # export-friendly version of nn.Hardswish()
+    @staticmethod
+    def forward(x):
+        # return x * F.hardsigmoid(x)  # for torchscript and CoreML
+        return x * F.hardtanh(x + 3, 0., 6.) / 6.  # for torchscript, CoreML and ONNX
+
+
+class MemoryEfficientSwish(nn.Module):
+    class F(torch.autograd.Function):
+        @staticmethod
+        def forward(ctx, x):
+            ctx.save_for_backward(x)
+            return x * torch.sigmoid(x)
+
+        @staticmethod
+        def backward(ctx, grad_output):
+            x = ctx.saved_tensors[0]
+            sx = torch.sigmoid(x)
+            return grad_output * (sx * (1 + x * (1 - sx)))
+
+    def forward(self, x):
+        return self.F.apply(x)
+
+
+# Mish https://github.com/digantamisra98/Mish --------------------------------------------------------------------------
+class Mish(nn.Module):
+    @staticmethod
+    def forward(x):
+        return x * F.softplus(x).tanh()
+
+
+class MemoryEfficientMish(nn.Module):
+    class F(torch.autograd.Function):
+        @staticmethod
+        def forward(ctx, x):
+            ctx.save_for_backward(x)
+            return x.mul(torch.tanh(F.softplus(x)))  # x * tanh(ln(1 + exp(x)))
+
+        @staticmethod
+        def backward(ctx, grad_output):
+            x = ctx.saved_tensors[0]
+            sx = torch.sigmoid(x)
+            fx = F.softplus(x).tanh()
+            return grad_output * (fx + x * sx * (1 - fx * fx))
+
+    def forward(self, x):
+        return self.F.apply(x)
+
+
+# FReLU https://arxiv.org/abs/2007.11824 -------------------------------------------------------------------------------
+class FReLU(nn.Module):
+    def __init__(self, c1, k=3):  # ch_in, kernel
+        super().__init__()
+        self.conv = nn.Conv2d(c1, c1, k, 1, 1, groups=c1, bias=False)
+        self.bn = nn.BatchNorm2d(c1)
+
+    def forward(self, x):
+        return torch.max(x, self.bn(self.conv(x)))

utils/autoanchor.py

@@ -0,0 +1,155 @@
+# Auto-anchor utils
+
+import numpy as np
+import torch
+import yaml
+from scipy.cluster.vq import kmeans
+from tqdm import tqdm
+
+from utils.general import colorstr
+
+
+def check_anchor_order(m):
+    # Check anchor order against stride order for YOLOv5 Detect() module m, and correct if necessary
+    a = m.anchor_grid.prod(-1).view(-1)  # anchor area
+    da = a[-1] - a[0]  # delta a
+    ds = m.stride[-1] - m.stride[0]  # delta s
+    if da.sign() != ds.sign():  # same order
+        print('Reversing anchor order')
+        m.anchors[:] = m.anchors.flip(0)
+        m.anchor_grid[:] = m.anchor_grid.flip(0)
+
+
+def check_anchors(dataset, model, thr=4.0, imgsz=640):
+    # Check anchor fit to data, recompute if necessary
+    prefix = colorstr('autoanchor: ')
+    print(f'\n{prefix}Analyzing anchors... ', end='')
+    m = model.module.model[-1] if hasattr(model, 'module') else model.model[-1]  # Detect()
+    shapes = imgsz * dataset.shapes / dataset.shapes.max(1, keepdims=True)
+    scale = np.random.uniform(0.9, 1.1, size=(shapes.shape[0], 1))  # augment scale
+    wh = torch.tensor(np.concatenate([l[:, 3:5] * s for s, l in zip(shapes * scale, dataset.labels)])).float()  # wh
+
+    def metric(k):  # compute metric
+        r = wh[:, None] / k[None]
+        x = torch.min(r, 1. / r).min(2)[0]  # ratio metric
+        best = x.max(1)[0]  # best_x
+        aat = (x > 1. / thr).float().sum(1).mean()  # anchors above threshold
+        bpr = (best > 1. / thr).float().mean()  # best possible recall
+        return bpr, aat
+
+    bpr, aat = metric(m.anchor_grid.clone().cpu().view(-1, 2))
+    print(f'anchors/target = {aat:.2f}, Best Possible Recall (BPR) = {bpr:.4f}', end='')
+    if bpr < 0.98:  # threshold to recompute
+        print('. Attempting to improve anchors, please wait...')
+        na = m.anchor_grid.numel() // 2  # number of anchors
+        new_anchors = kmean_anchors(dataset, n=na, img_size=imgsz, thr=thr, gen=1000, verbose=False)
+        new_bpr = metric(new_anchors.reshape(-1, 2))[0]
+        if new_bpr > bpr:  # replace anchors
+            new_anchors = torch.tensor(new_anchors, device=m.anchors.device).type_as(m.anchors)
+            m.anchor_grid[:] = new_anchors.clone().view_as(m.anchor_grid)  # for inference
+            m.anchors[:] = new_anchors.clone().view_as(m.anchors) / m.stride.to(m.anchors.device).view(-1, 1, 1)  # loss
+            check_anchor_order(m)
+            print(f'{prefix}New anchors saved to model. Update model *.yaml to use these anchors in the future.')
+        else:
+            print(f'{prefix}Original anchors better than new anchors. Proceeding with original anchors.')
+    print('')  # newline
+
+
+def kmean_anchors(path='./data/coco128.yaml', n=9, img_size=640, thr=4.0, gen=1000, verbose=True):
+    """ Creates kmeans-evolved anchors from training dataset
+
+        Arguments:
+            path: path to dataset *.yaml, or a loaded dataset
+            n: number of anchors
+            img_size: image size used for training
+            thr: anchor-label wh ratio threshold hyperparameter hyp['anchor_t'] used for training, default=4.0
+            gen: generations to evolve anchors using genetic algorithm
+            verbose: print all results
+
+        Return:
+            k: kmeans evolved anchors
+
+        Usage:
+            from utils.autoanchor import *; _ = kmean_anchors()
+    """
+    thr = 1. / thr
+    prefix = colorstr('autoanchor: ')
+
+    def metric(k, wh):  # compute metrics
+        r = wh[:, None] / k[None]
+        x = torch.min(r, 1. / r).min(2)[0]  # ratio metric
+        # x = wh_iou(wh, torch.tensor(k))  # iou metric
+        return x, x.max(1)[0]  # x, best_x
+
+    def anchor_fitness(k):  # mutation fitness
+        _, best = metric(torch.tensor(k, dtype=torch.float32), wh)
+        return (best * (best > thr).float()).mean()  # fitness
+
+    def print_results(k):
+        k = k[np.argsort(k.prod(1))]  # sort small to large
+        x, best = metric(k, wh0)
+        bpr, aat = (best > thr).float().mean(), (x > thr).float().mean() * n  # best possible recall, anch > thr
+        print(f'{prefix}thr={thr:.2f}: {bpr:.4f} best possible recall, {aat:.2f} anchors past thr')
+        print(f'{prefix}n={n}, img_size={img_size}, metric_all={x.mean():.3f}/{best.mean():.3f}-mean/best, '
+              f'past_thr={x[x > thr].mean():.3f}-mean: ', end='')
+        for i, x in enumerate(k):
+            print('%i,%i' % (round(x[0]), round(x[1])), end=',  ' if i < len(k) - 1 else '\n')  # use in *.cfg
+        return k
+
+    if isinstance(path, str):  # *.yaml file
+        with open(path) as f:
+            data_dict = yaml.load(f, Loader=yaml.SafeLoader)  # model dict
+        from utils.datasets import LoadImagesAndLabels
+        dataset = LoadImagesAndLabels(data_dict['train'], augment=True, rect=True)
+    else:
+        dataset = path  # dataset
+
+    # Get label wh
+    shapes = img_size * dataset.shapes / dataset.shapes.max(1, keepdims=True)
+    wh0 = np.concatenate([l[:, 3:5] * s for s, l in zip(shapes, dataset.labels)])  # wh
+
+    # Filter
+    i = (wh0 < 3.0).any(1).sum()
+    if i:
+        print(f'{prefix}WARNING: Extremely small objects found. {i} of {len(wh0)} labels are < 3 pixels in size.')
+    wh = wh0[(wh0 >= 2.0).any(1)]  # filter > 2 pixels
+    # wh = wh * (np.random.rand(wh.shape[0], 1) * 0.9 + 0.1)  # multiply by random scale 0-1
+
+    # Kmeans calculation
+    print(f'{prefix}Running kmeans for {n} anchors on {len(wh)} points...')
+    s = wh.std(0)  # sigmas for whitening
+    k, dist = kmeans(wh / s, n, iter=30)  # points, mean distance
+    k *= s
+    wh = torch.tensor(wh, dtype=torch.float32)  # filtered
+    wh0 = torch.tensor(wh0, dtype=torch.float32)  # unfiltered
+    k = print_results(k)
+
+    # Plot
+    # k, d = [None] * 20, [None] * 20
+    # for i in tqdm(range(1, 21)):
+    #     k[i-1], d[i-1] = kmeans(wh / s, i)  # points, mean distance
+    # fig, ax = plt.subplots(1, 2, figsize=(14, 7), tight_layout=True)
+    # ax = ax.ravel()
+    # ax[0].plot(np.arange(1, 21), np.array(d) ** 2, marker='.')
+    # fig, ax = plt.subplots(1, 2, figsize=(14, 7))  # plot wh
+    # ax[0].hist(wh[wh[:, 0]<100, 0],400)
+    # ax[1].hist(wh[wh[:, 1]<100, 1],400)
+    # fig.savefig('wh.png', dpi=200)
+
+    # Evolve
+    npr = np.random
+    f, sh, mp, s = anchor_fitness(k), k.shape, 0.9, 0.1  # fitness, generations, mutation prob, sigma
+    pbar = tqdm(range(gen), desc=f'{prefix}Evolving anchors with Genetic Algorithm:')  # progress bar
+    for _ in pbar:
+        v = np.ones(sh)
+        while (v == 1).all():  # mutate until a change occurs (prevent duplicates)
+            v = ((npr.random(sh) < mp) * npr.random() * npr.randn(*sh) * s + 1).clip(0.3, 3.0)
+        kg = (k.copy() * v).clip(min=2.0)
+        fg = anchor_fitness(kg)
+        if fg > f:
+            f, k = fg, kg.copy()
+            pbar.desc = f'{prefix}Evolving anchors with Genetic Algorithm: fitness = {f:.4f}'
+            if verbose:
+                print_results(k)
+
+    return print_results(k)

utils/aws/mime.sh

@@ -0,0 +1,26 @@
+# AWS EC2 instance startup 'MIME' script https://aws.amazon.com/premiumsupport/knowledge-center/execute-user-data-ec2/
+# This script will run on every instance restart, not only on first start
+# --- DO NOT COPY ABOVE COMMENTS WHEN PASTING INTO USERDATA ---
+
+Content-Type: multipart/mixed; boundary="//"
+MIME-Version: 1.0
+
+--//
+Content-Type: text/cloud-config; charset="us-ascii"
+MIME-Version: 1.0
+Content-Transfer-Encoding: 7bit
+Content-Disposition: attachment; filename="cloud-config.txt"
+
+#cloud-config
+cloud_final_modules:
+- [scripts-user, always]
+
+--//
+Content-Type: text/x-shellscript; charset="us-ascii"
+MIME-Version: 1.0
+Content-Transfer-Encoding: 7bit
+Content-Disposition: attachment; filename="userdata.txt"
+
+#!/bin/bash
+# --- paste contents of userdata.sh here ---
+--//

utils/aws/resume.py

@@ -0,0 +1,37 @@
+# Resume all interrupted trainings in yolov5/ dir including DDP trainings
+# Usage: $ python utils/aws/resume.py
+
+import os
+import sys
+from pathlib import Path
+
+import torch
+import yaml
+
+sys.path.append('./')  # to run '$ python *.py' files in subdirectories
+
+port = 0  # --master_port
+path = Path('').resolve()
+for last in path.rglob('*/**/last.pt'):
+    ckpt = torch.load(last)
+    if ckpt['optimizer'] is None:
+        continue
+
+    # Load opt.yaml
+    with open(last.parent.parent / 'opt.yaml') as f:
+        opt = yaml.load(f, Loader=yaml.SafeLoader)
+
+    # Get device count
+    d = opt['device'].split(',')  # devices
+    nd = len(d)  # number of devices
+    ddp = nd > 1 or (nd == 0 and torch.cuda.device_count() > 1)  # distributed data parallel
+
+    if ddp:  # multi-GPU
+        port += 1
+        cmd = f'python -m torch.distributed.launch --nproc_per_node {nd} --master_port {port} train.py --resume {last}'
+    else:  # single-GPU
+        cmd = f'python train.py --resume {last}'
+
+    cmd += ' > /dev/null 2>&1 &'  # redirect output to dev/null and run in daemon thread
+    print(cmd)
+    os.system(cmd)

utils/aws/userdata.sh

@@ -0,0 +1,27 @@
+#!/bin/bash
+# AWS EC2 instance startup script https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/user-data.html
+# This script will run only once on first instance start (for a re-start script see mime.sh)
+# /home/ubuntu (ubuntu) or /home/ec2-user (amazon-linux) is working dir
+# Use >300 GB SSD
+
+cd home/ubuntu
+if [ ! -d yolov5 ]; then
+  echo "Running first-time script." # install dependencies, download COCO, pull Docker
+  git clone https://github.com/ultralytics/yolov5 && sudo chmod -R 777 yolov5
+  cd yolov5
+  bash data/scripts/get_coco.sh && echo "Data done." &
+  sudo docker pull ultralytics/yolov5:latest && echo "Docker done." &
+  python -m pip install --upgrade pip && pip install -r requirements.txt && python detect.py && echo "Requirements done." &
+  wait && echo "All tasks done." # finish background tasks
+else
+  echo "Running re-start script." # resume interrupted runs
+  i=0
+  list=$(sudo docker ps -qa) # container list i.e. $'one\ntwo\nthree\nfour'
+  while IFS= read -r id; do
+    ((i++))
+    echo "restarting container $i: $id"
+    sudo docker start $id
+    # sudo docker exec -it $id python train.py --resume # single-GPU
+    sudo docker exec -d $id python utils/aws/resume.py # multi-scenario
+  done <<<"$list"
+fi

utils/cv_puttext.py

@@ -0,0 +1,22 @@
+import cv2
+import numpy as np
+from PIL import Image, ImageDraw, ImageFont
+
+def cv2ImgAddText(img, text, left, top, textColor=(0, 255, 0), textSize=20):
+    if (isinstance(img, np.ndarray)):  #判断是否OpenCV图片类型
+        img = Image.fromarray(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
+    draw = ImageDraw.Draw(img)
+    fontText = ImageFont.truetype(
+        "fonts/platech.ttf", textSize, encoding="utf-8")
+    draw.text((left, top), text, textColor, font=fontText)
+    return cv2.cvtColor(np.asarray(img), cv2.COLOR_RGB2BGR)
+
+if __name__ == '__main__':
+    imgPath = "result.jpg"
+    img = cv2.imread(imgPath)
+    
+    saveImg = cv2ImgAddText(img, '中国加油！', 50, 100, (255, 0, 0), 50)
+    
+    # cv2.imshow('display',saveImg)
+    cv2.imwrite('save.jpg',saveImg)
+    # cv2.waitKey()

utils/face_datasets.py

@@ -0,0 +1,843 @@
+import glob
+import logging
+import math
+import os
+import random
+import shutil
+import time
+from itertools import repeat
+from multiprocessing.pool import ThreadPool
+from pathlib import Path
+from threading import Thread
+
+import cv2
+import numpy as np
+import torch
+from PIL import Image, ExifTags
+from torch.utils.data import Dataset
+from tqdm import tqdm
+
+from utils.general import xyxy2xywh, xywh2xyxy, clean_str
+from utils.torch_utils import torch_distributed_zero_first
+
+
+# Parameters
+help_url = 'https://github.com/ultralytics/yolov5/wiki/Train-Custom-Data'
+img_formats = ['bmp', 'jpg', 'jpeg', 'png', 'tif', 'tiff', 'dng']  # acceptable image suffixes
+vid_formats = ['mov', 'avi', 'mp4', 'mpg', 'mpeg', 'm4v', 'wmv', 'mkv']  # acceptable video suffixes
+logger = logging.getLogger(__name__)
+
+# Get orientation exif tag
+for orientation in ExifTags.TAGS.keys():
+    if ExifTags.TAGS[orientation] == 'Orientation':
+        break
+
+def get_hash(files):
+    # Returns a single hash value of a list of files
+    return sum(os.path.getsize(f) for f in files if os.path.isfile(f))
+
+def img2label_paths(img_paths):
+    # Define label paths as a function of image paths
+    sa, sb = os.sep + 'images' + os.sep, os.sep + 'labels' + os.sep  # /images/, /labels/ substrings
+    return [x.replace(sa, sb, 1).replace('.' + x.split('.')[-1], '.txt') for x in img_paths]
+
+def exif_size(img):
+    # Returns exif-corrected PIL size
+    s = img.size  # (width, height)
+    try:
+        rotation = dict(img._getexif().items())[orientation]
+        if rotation == 6:  # rotation 270
+            s = (s[1], s[0])
+        elif rotation == 8:  # rotation 90
+            s = (s[1], s[0])
+    except:
+        pass
+
+    return s
+
+def create_dataloader(path, imgsz, batch_size, stride, opt, hyp=None, augment=False, cache=False, pad=0.0, rect=False,
+                      rank=-1, world_size=1, workers=8, image_weights=False, quad=False, prefix=''):
+    # Make sure only the first process in DDP process the dataset first, and the following others can use the cache
+    with torch_distributed_zero_first(rank):
+        dataset = LoadFaceImagesAndLabels(path, imgsz, batch_size,
+                                      augment=augment,  # augment images
+                                      hyp=hyp,  # augmentation hyperparameters
+                                      rect=rect,  # rectangular training
+                                      cache_images=cache,
+                                      single_cls=opt.single_cls,
+                                      stride=int(stride),
+                                      pad=pad,
+                                      image_weights=image_weights,
+                                    )
+
+    batch_size = min(batch_size, len(dataset))
+    nw = min([os.cpu_count() // world_size, batch_size if batch_size > 1 else 0, workers])  # number of workers
+    sampler = torch.utils.data.distributed.DistributedSampler(dataset) if rank != -1 else None
+    loader = torch.utils.data.DataLoader if image_weights else InfiniteDataLoader
+    # Use torch.utils.data.DataLoader() if dataset.properties will update during training else InfiniteDataLoader()
+    dataloader = loader(dataset,
+                        batch_size=batch_size,
+                        num_workers=nw,
+                        sampler=sampler,
+                        pin_memory=True,
+                        collate_fn=LoadFaceImagesAndLabels.collate_fn4 if quad else LoadFaceImagesAndLabels.collate_fn)
+    return dataloader, dataset
+class InfiniteDataLoader(torch.utils.data.dataloader.DataLoader):
+    """ Dataloader that reuses workers
+
+    Uses same syntax as vanilla DataLoader
+    """
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)
+        object.__setattr__(self, 'batch_sampler', _RepeatSampler(self.batch_sampler))
+        self.iterator = super().__iter__()
+
+    def __len__(self):
+        return len(self.batch_sampler.sampler)
+
+    def __iter__(self):
+        for i in range(len(self)):
+            yield next(self.iterator)
+class _RepeatSampler(object):
+    """ Sampler that repeats forever
+
+    Args:
+        sampler (Sampler)
+    """
+
+    def __init__(self, sampler):
+        self.sampler = sampler
+
+    def __iter__(self):
+        while True:
+            yield from iter(self.sampler)
+
+class LoadFaceImagesAndLabels(Dataset):  # for training/testing
+    def __init__(self, path, img_size=640, batch_size=16, augment=False, hyp=None, rect=False, image_weights=False,
+                 cache_images=False, single_cls=False, stride=32, pad=0.0, rank=-1):
+        self.img_size = img_size
+        self.augment = augment
+        self.hyp = hyp
+        self.image_weights = image_weights
+        self.rect = False if image_weights else rect
+        self.mosaic = self.augment and not self.rect  # load 4 images at a time into a mosaic (only during training)
+        self.mosaic_border = [-img_size // 2, -img_size // 2]
+        self.stride = stride
+
+        try:
+            f = []  # image files
+            for p in path if isinstance(path, list) else [path]:
+                p = Path(p)  # os-agnostic
+                if p.is_dir():  # dir
+                    f += glob.glob(str(p / '**' / '*.*'), recursive=True)
+                elif p.is_file():  # file
+                    with open(p, 'r') as t:
+                        t = t.read().strip().splitlines()
+                        parent = str(p.parent) + os.sep
+                        f += [x.replace('./', parent) if x.startswith('./') else x for x in t]  # local to global path
+                else:
+                    raise Exception('%s does not exist' % p)
+            self.img_files = sorted([x.replace('/', os.sep) for x in f if x.split('.')[-1].lower() in img_formats])
+            assert self.img_files, 'No images found'
+        except Exception as e:
+            raise Exception('Error loading data from %s: %s\nSee %s' % (path, e, help_url))
+
+        # Check cache
+        self.label_files = img2label_paths(self.img_files)  # labels
+        cache_path = Path(self.label_files[0]).parent.with_suffix('.cache')  # cached labels
+        if cache_path.is_file():
+            cache = torch.load(cache_path)  # load
+            if cache['hash'] != get_hash(self.label_files + self.img_files) or 'results' not in cache:  # changed
+                cache = self.cache_labels(cache_path)  # re-cache
+        else:
+            cache = self.cache_labels(cache_path)  # cache
+
+        # Display cache
+        [nf, nm, ne, nc, n] = cache.pop('results')  # found, missing, empty, corrupted, total
+        desc = f"Scanning '{cache_path}' for images and labels... {nf} found, {nm} missing, {ne} empty, {nc} corrupted"
+        tqdm(None, desc=desc, total=n, initial=n)
+        assert nf > 0 or not augment, f'No labels found in {cache_path}. Can not train without labels. See {help_url}'
+
+        # Read cache
+        cache.pop('hash')  # remove hash
+        labels, shapes = zip(*cache.values())
+        self.labels = list(labels)
+        self.shapes = np.array(shapes, dtype=np.float64)
+        self.img_files = list(cache.keys())  # update
+        self.label_files = img2label_paths(cache.keys())  # update
+        if single_cls:
+            for x in self.labels:
+                x[:, 0] = 0
+
+        n = len(shapes)  # number of images
+        bi = np.floor(np.arange(n) / batch_size).astype(np.int)  # batch index
+        nb = bi[-1] + 1  # number of batches
+        self.batch = bi  # batch index of image
+        self.n = n
+        self.indices = range(n)
+
+        # Rectangular Training
+        if self.rect:
+            # Sort by aspect ratio
+            s = self.shapes  # wh
+            ar = s[:, 1] / s[:, 0]  # aspect ratio
+            irect = ar.argsort()
+            self.img_files = [self.img_files[i] for i in irect]
+            self.label_files = [self.label_files[i] for i in irect]
+            self.labels = [self.labels[i] for i in irect]
+            self.shapes = s[irect]  # wh
+            ar = ar[irect]
+
+            # Set training image shapes
+            shapes = [[1, 1]] * nb
+            for i in range(nb):
+                ari = ar[bi == i]
+                mini, maxi = ari.min(), ari.max()
+                if maxi < 1:
+                    shapes[i] = [maxi, 1]
+                elif mini > 1:
+                    shapes[i] = [1, 1 / mini]
+
+            self.batch_shapes = np.ceil(np.array(shapes) * img_size / stride + pad).astype(np.int) * stride
+
+        # Cache images into memory for faster training (WARNING: large datasets may exceed system RAM)
+        self.imgs = [None] * n
+        if cache_images:
+            gb = 0  # Gigabytes of cached images
+            self.img_hw0, self.img_hw = [None] * n, [None] * n
+            results = ThreadPool(8).imap(lambda x: load_image(*x), zip(repeat(self), range(n)))  # 8 threads
+            pbar = tqdm(enumerate(results), total=n)
+            for i, x in pbar:
+                self.imgs[i], self.img_hw0[i], self.img_hw[i] = x  # img, hw_original, hw_resized = load_image(self, i)
+                gb += self.imgs[i].nbytes
+                pbar.desc = 'Caching images (%.1fGB)' % (gb / 1E9)
+
+    def cache_labels(self, path=Path('./labels.cache')):
+        # Cache dataset labels, check images and read shapes
+        x = {}  # dict
+        nm, nf, ne, nc = 0, 0, 0, 0  # number missing, found, empty, duplicate
+        pbar = tqdm(zip(self.img_files, self.label_files), desc='Scanning images', total=len(self.img_files))
+        for i, (im_file, lb_file) in enumerate(pbar):
+            try:
+                # verify images
+                im = Image.open(im_file)
+                im.verify()  # PIL verify
+                shape = exif_size(im)  # image size
+                assert (shape[0] > 9) & (shape[1] > 9), 'image size <10 pixels'
+
+                # verify labels
+                if os.path.isfile(lb_file):
+                    nf += 1  # label found
+                    with open(lb_file, 'r') as f:
+                        l = np.array([x.split() for x in f.read().strip().splitlines()], dtype=np.float32)  # labels
+                    if len(l):
+                        assert l.shape[1] == 13, 'labels require 13 columns each'
+                        assert (l >= -1).all(), 'negative labels'
+                        assert (l[:, 1:] <= 1).all(), 'non-normalized or out of bounds coordinate labels'
+                        assert np.unique(l, axis=0).shape[0] == l.shape[0], 'duplicate labels'
+                    else:
+                        ne += 1  # label empty
+                        l = np.zeros((0, 13), dtype=np.float32)
+                else:
+                    nm += 1  # label missing
+                    l = np.zeros((0, 13), dtype=np.float32)
+                x[im_file] = [l, shape]
+            except Exception as e:
+                nc += 1
+                print('WARNING: Ignoring corrupted image and/or label %s: %s' % (im_file, e))
+
+            pbar.desc = f"Scanning '{path.parent / path.stem}' for images and labels... " \
+                        f"{nf} found, {nm} missing, {ne} empty, {nc} corrupted"
+
+        if nf == 0:
+            print(f'WARNING: No labels found in {path}. See {help_url}')
+
+        x['hash'] = get_hash(self.label_files + self.img_files)
+        x['results'] = [nf, nm, ne, nc, i + 1]
+        torch.save(x, path)  # save for next time
+        logging.info(f"New cache created: {path}")
+        return x
+
+    def __len__(self):
+        return len(self.img_files)
+
+    # def __iter__(self):
+    #     self.count = -1
+    #     print('ran dataset iter')
+    #     #self.shuffled_vector = np.random.permutation(self.nF) if self.augment else np.arange(self.nF)
+    #     return self
+
+    def __getitem__(self, index):
+        index = self.indices[index]  # linear, shuffled, or image_weights
+
+        hyp = self.hyp
+        mosaic = self.mosaic and random.random() < hyp['mosaic']
+        if mosaic:
+            # Load mosaic
+            img, labels = load_mosaic_face(self, index)
+            shapes = None
+
+            # MixUp https://arxiv.org/pdf/1710.09412.pdf
+            if random.random() < hyp['mixup']:
+                img2, labels2 = load_mosaic_face(self, random.randint(0, self.n - 1))
+                r = np.random.beta(8.0, 8.0)  # mixup ratio, alpha=beta=8.0
+                img = (img * r + img2 * (1 - r)).astype(np.uint8)
+                labels = np.concatenate((labels, labels2), 0)
+
+        else:
+            # Load image
+            img, (h0, w0), (h, w) = load_image(self, index)
+
+            # Letterbox
+            shape = self.batch_shapes[self.batch[index]] if self.rect else self.img_size  # final letterboxed shape
+            img, ratio, pad = letterbox(img, shape, auto=False, scaleup=self.augment)
+            shapes = (h0, w0), ((h / h0, w / w0), pad)  # for COCO mAP rescaling
+
+            # Load labels
+            labels = []
+            x = self.labels[index]
+            if x.size > 0:
+                # Normalized xywh to pixel xyxy format
+                labels = x.copy()
+                labels[:, 1] = ratio[0] * w * (x[:, 1] - x[:, 3] / 2) + pad[0]  # pad width
+                labels[:, 2] = ratio[1] * h * (x[:, 2] - x[:, 4] / 2) + pad[1]  # pad height
+                labels[:, 3] = ratio[0] * w * (x[:, 1] + x[:, 3] / 2) + pad[0]
+                labels[:, 4] = ratio[1] * h * (x[:, 2] + x[:, 4] / 2) + pad[1]
+
+                #labels[:, 5] = ratio[0] * w * x[:, 5] + pad[0]  # pad width
+                labels[:, 5] = np.array(x[:, 5] > 0, dtype=np.int32) * (ratio[0] * w * x[:, 5] + pad[0]) + (
+                    np.array(x[:, 5] > 0, dtype=np.int32) - 1)
+                labels[:, 6] = np.array(x[:, 6] > 0, dtype=np.int32) * (ratio[1] * h * x[:, 6] + pad[1]) + (
+                    np.array(x[:, 6] > 0, dtype=np.int32) - 1)
+                labels[:, 7] = np.array(x[:, 7] > 0, dtype=np.int32) * (ratio[0] * w * x[:, 7] + pad[0]) + (
+                    np.array(x[:, 7] > 0, dtype=np.int32) - 1)
+                labels[:, 8] = np.array(x[:, 8] > 0, dtype=np.int32) * (ratio[1] * h * x[:, 8] + pad[1]) + (
+                    np.array(x[:, 8] > 0, dtype=np.int32) - 1)
+                labels[:, 9] = np.array(x[:, 5] > 0, dtype=np.int32) * (ratio[0] * w * x[:, 9] + pad[0]) + (
+                    np.array(x[:, 9] > 0, dtype=np.int32) - 1)
+                labels[:, 10] = np.array(x[:, 5] > 0, dtype=np.int32) * (ratio[1] * h * x[:, 10] + pad[1]) + (
+                    np.array(x[:, 10] > 0, dtype=np.int32) - 1)
+                labels[:, 11] = np.array(x[:, 11] > 0, dtype=np.int32) * (ratio[0] * w * x[:, 11] + pad[0]) + (
+                    np.array(x[:, 11] > 0, dtype=np.int32) - 1)
+                labels[:, 12] = np.array(x[:, 12] > 0, dtype=np.int32) * (ratio[1] * h * x[:, 12] + pad[1]) + (
+                    np.array(x[:, 12] > 0, dtype=np.int32) - 1)
+                # labels[:, 13] = np.array(x[:, 13] > 0, dtype=np.int32) * (ratio[0] * w * x[:, 13] + pad[0]) + (
+                #     np.array(x[:, 13] > 0, dtype=np.int32) - 1)
+                # labels[:, 14] = np.array(x[:, 14] > 0, dtype=np.int32) * (ratio[1] * h * x[:, 14] + pad[1]) + (
+                #     np.array(x[:, 14] > 0, dtype=np.int32) - 1)
+
+        if self.augment:
+            # Augment imagespace
+            if not mosaic:
+                img, labels = random_perspective(img, labels,
+                                                 degrees=hyp['degrees'],
+                                                 translate=hyp['translate'],
+                                                 scale=hyp['scale'],
+                                                 shear=hyp['shear'],
+                                                 perspective=hyp['perspective'])
+
+            # Augment colorspace
+            augment_hsv(img, hgain=hyp['hsv_h'], sgain=hyp['hsv_s'], vgain=hyp['hsv_v'])
+
+            # Apply cutouts
+            # if random.random() < 0.9:
+            #     labels = cutout(img, labels)
+
+        nL = len(labels)  # number of labels
+        if nL:
+            labels[:, 1:5] = xyxy2xywh(labels[:, 1:5])  # convert xyxy to xywh
+            labels[:, [2, 4]] /= img.shape[0]  # normalized height 0-1
+            labels[:, [1, 3]] /= img.shape[1]  # normalized width 0-1
+
+            labels[:, [5, 7, 9, 11]] /= img.shape[1]  # normalized landmark x 0-1
+            labels[:, [5, 7, 9, 11]] = np.where(labels[:, [5, 7, 9, 11]] < 0, -1, labels[:, [5, 7, 9, 11]])
+            labels[:, [6, 8, 10, 12]] /= img.shape[0]  # normalized landmark y 0-1
+            labels[:, [6, 8, 10, 12]] = np.where(labels[:, [6, 8, 10, 12]] < 0, -1, labels[:, [6, 8, 10, 12]])
+
+        if self.augment:
+            # flip up-down
+            if random.random() < hyp['flipud']:
+                img = np.flipud(img)
+                if nL:
+                    labels[:, 2] = 1 - labels[:, 2]
+
+                    labels[:, 6] = np.where(labels[:,6] < 0, -1, 1 - labels[:, 6])
+                    labels[:, 8] = np.where(labels[:, 8] < 0, -1, 1 - labels[:, 8])
+                    labels[:, 10] = np.where(labels[:, 10] < 0, -1, 1 - labels[:, 10])
+                    labels[:, 12] = np.where(labels[:, 12] < 0, -1, 1 - labels[:, 12])
+                    # labels[:, 14] = np.where(labels[:, 14] < 0, -1, 1 - labels[:, 14])
+
+            # flip left-right
+            if random.random() < hyp['fliplr']:
+                img = np.fliplr(img)
+                if nL:
+                    labels[:, 1] = 1 - labels[:, 1]
+
+                    labels[:, 5] = np.where(labels[:, 5] < 0, -1, 1 - labels[:, 5])
+                    labels[:, 7] = np.where(labels[:, 7] < 0, -1, 1 - labels[:, 7])
+                    labels[:, 9] = np.where(labels[:, 9] < 0, -1, 1 - labels[:, 9])
+                    labels[:, 11] = np.where(labels[:, 11] < 0, -1, 1 - labels[:, 11])
+                    # labels[:, 13] = np.where(labels[:, 13] < 0, -1, 1 - labels[:, 13])
+
+                    #左右镜像的时候，关键点应该交换位置，不然的话顺序就错了
+                    left_top = np.copy(labels[:, [5, 6]])
+                    left_bottom = np.copy(labels[:, [9, 10]])
+                    labels[:, [5, 6]] = labels[:, [7, 8]]
+                    labels[:, [7, 8]] = left_top
+                    labels[:, [9, 10]] = labels[:, [11, 12]]
+                    labels[:, [11, 12]] = left_bottom
+
+
+
+                    # eye_left = np.copy(labels[:, [5, 6]])
+                    # mouth_left = np.copy(labels[:, [11, 12]])
+                    # labels[:, [5, 6]] = labels[:, [7, 8]]
+                    # labels[:, [7, 8]] = eye_left
+                    # labels[:, [11, 12]] = labels[:, [13, 14]]
+                    # labels[:, [13, 14]] = mouth_left
+
+        labels_out = torch.zeros((nL, 14))
+        if nL:
+            labels_out[:, 1:] = torch.from_numpy(labels)
+            #showlabels(img, labels[:, 1:5], labels[:, 5:13])
+
+        # Convert
+        img = img[:, :, ::-1].transpose(2, 0, 1)  # BGR to RGB, to 3x416x416
+        img = np.ascontiguousarray(img)
+        #print(index, '   --- labels_out: ', labels_out)
+        #if nL:
+            #print( ' : landmarks : ', torch.max(labels_out[:, 5:13]), '  ---   ', torch.min(labels_out[:, 5:13]))
+        return torch.from_numpy(img), labels_out, self.img_files[index], shapes
+
+    @staticmethod
+    def collate_fn(batch):
+        img, label, path, shapes = zip(*batch)  # transposed
+        for i, l in enumerate(label):
+            l[:, 0] = i  # add target image index for build_targets()
+        return torch.stack(img, 0), torch.cat(label, 0), path, shapes
+
+
+def showlabels(img, boxs, landmarks):
+    for box in boxs:
+        x,y,w,h = box[0] * img.shape[1], box[1] * img.shape[0], box[2] * img.shape[1], box[3] * img.shape[0]
+        #cv2.rectangle(image, (x,y), (x+w,y+h), (0,255,0), 2)
+        cv2.rectangle(img, (int(x - w/2), int(y - h/2)), (int(x + w/2), int(y + h/2)), (0, 255, 0), 2)
+
+    for landmark in landmarks:
+        #cv2.circle(img,(60,60),30,(0,0,255))
+        for i in range(4):
+            cv2.circle(img, (int(landmark[2*i] * img.shape[1]), int(landmark[2*i+1]*img.shape[0])), 3 ,(0,0,255), -1)
+    cv2.imshow('test', img)
+    cv2.waitKey(0)
+
+
+def load_mosaic_face(self, index):
+    # loads images in a mosaic
+    labels4 = []
+    s = self.img_size
+    yc, xc = [int(random.uniform(-x, 2 * s + x)) for x in self.mosaic_border]  # mosaic center x, y
+    indices = [index] + [self.indices[random.randint(0, self.n - 1)] for _ in range(3)]  # 3 additional image indices
+    for i, index in enumerate(indices):
+        # Load image
+        img, _, (h, w) = load_image(self, index)
+
+        # place img in img4
+        if i == 0:  # top left
+            img4 = np.full((s * 2, s * 2, img.shape[2]), 114, dtype=np.uint8)  # base image with 4 tiles
+            x1a, y1a, x2a, y2a = max(xc - w, 0), max(yc - h, 0), xc, yc  # xmin, ymin, xmax, ymax (large image)
+            x1b, y1b, x2b, y2b = w - (x2a - x1a), h - (y2a - y1a), w, h  # xmin, ymin, xmax, ymax (small image)
+        elif i == 1:  # top right
+            x1a, y1a, x2a, y2a = xc, max(yc - h, 0), min(xc + w, s * 2), yc
+            x1b, y1b, x2b, y2b = 0, h - (y2a - y1a), min(w, x2a - x1a), h
+        elif i == 2:  # bottom left
+            x1a, y1a, x2a, y2a = max(xc - w, 0), yc, xc, min(s * 2, yc + h)
+            x1b, y1b, x2b, y2b = w - (x2a - x1a), 0, w, min(y2a - y1a, h)
+        elif i == 3:  # bottom right
+            x1a, y1a, x2a, y2a = xc, yc, min(xc + w, s * 2), min(s * 2, yc + h)
+            x1b, y1b, x2b, y2b = 0, 0, min(w, x2a - x1a), min(y2a - y1a, h)
+
+        img4[y1a:y2a, x1a:x2a] = img[y1b:y2b, x1b:x2b]  # img4[ymin:ymax, xmin:xmax]
+        padw = x1a - x1b
+        padh = y1a - y1b
+
+        # Labels
+        x = self.labels[index]
+        labels = x.copy()
+        if x.size > 0:  # Normalized xywh to pixel xyxy format
+            #box, x1,y1,x2,y2
+            labels[:, 1] = w * (x[:, 1] - x[:, 3] / 2) + padw
+            labels[:, 2] = h * (x[:, 2] - x[:, 4] / 2) + padh
+            labels[:, 3] = w * (x[:, 1] + x[:, 3] / 2) + padw
+            labels[:, 4] = h * (x[:, 2] + x[:, 4] / 2) + padh
+            #10 landmarks
+
+            labels[:, 5] = np.array(x[:, 5] > 0, dtype=np.int32) * (w * x[:, 5] + padw) + (np.array(x[:, 5] > 0, dtype=np.int32) - 1)
+            labels[:, 6] = np.array(x[:, 6] > 0, dtype=np.int32) * (h * x[:, 6] + padh) + (np.array(x[:, 6] > 0, dtype=np.int32) - 1)
+            labels[:, 7] = np.array(x[:, 7] > 0, dtype=np.int32) * (w * x[:, 7] + padw) + (np.array(x[:, 7] > 0, dtype=np.int32) - 1)
+            labels[:, 8] = np.array(x[:, 8] > 0, dtype=np.int32) * (h * x[:, 8] + padh) + (np.array(x[:, 8] > 0, dtype=np.int32) - 1)
+            labels[:, 9] = np.array(x[:, 9] > 0, dtype=np.int32) * (w * x[:, 9] + padw) + (np.array(x[:, 9] > 0, dtype=np.int32) - 1)
+            labels[:, 10] = np.array(x[:, 10] > 0, dtype=np.int32) * (h * x[:, 10] + padh) + (np.array(x[:, 10] > 0, dtype=np.int32) - 1)
+            labels[:, 11] = np.array(x[:, 11] > 0, dtype=np.int32) * (w * x[:, 11] + padw) + (np.array(x[:, 11] > 0, dtype=np.int32) - 1)
+            labels[:, 12] = np.array(x[:, 12] > 0, dtype=np.int32) * (h * x[:, 12] + padh) + (np.array(x[:, 12] > 0, dtype=np.int32) - 1)
+            # labels[:, 13] = np.array(x[:, 13] > 0, dtype=np.int32) * (w * x[:, 13] + padw) + (np.array(x[:, 13] > 0, dtype=np.int32) - 1)
+            # labels[:, 14] = np.array(x[:, 14] > 0, dtype=np.int32) * (h * x[:, 14] + padh) + (np.array(x[:, 14] > 0, dtype=np.int32) - 1)
+        labels4.append(labels)
+
+    # Concat/clip labels
+    if len(labels4):
+        labels4 = np.concatenate(labels4, 0)
+        np.clip(labels4[:, 1:5], 0, 2 * s, out=labels4[:, 1:5])  # use with random_perspective
+        # img4, labels4 = replicate(img4, labels4)  # replicate
+
+        #landmarks
+        labels4[:, 5:] = np.where(labels4[:, 5:] < 0, -1, labels4[:, 5:])
+        labels4[:, 5:] = np.where(labels4[:, 5:] > 2 * s, -1, labels4[:, 5:])
+
+        labels4[:, 5] = np.where(labels4[:, 6] == -1, -1, labels4[:, 5])
+        labels4[:, 6] = np.where(labels4[:, 5] == -1, -1, labels4[:, 6])
+
+        labels4[:, 7] = np.where(labels4[:, 8] == -1, -1, labels4[:, 7])
+        labels4[:, 8] = np.where(labels4[:, 7] == -1, -1, labels4[:, 8])
+
+        labels4[:, 9] = np.where(labels4[:, 10] == -1, -1, labels4[:, 9])
+        labels4[:, 10] = np.where(labels4[:, 9] == -1, -1, labels4[:, 10])
+
+        labels4[:, 11] = np.where(labels4[:, 12] == -1, -1, labels4[:, 11])
+        labels4[:, 12] = np.where(labels4[:, 11] == -1, -1, labels4[:, 12])
+
+        # labels4[:, 13] = np.where(labels4[:, 14] == -1, -1, labels4[:, 13])
+        # labels4[:, 14] = np.where(labels4[:, 13] == -1, -1, labels4[:, 14])
+
+    # Augment
+    img4, labels4 = random_perspective(img4, labels4,
+                                       degrees=self.hyp['degrees'],
+                                       translate=self.hyp['translate'],
+                                       scale=self.hyp['scale'],
+                                       shear=self.hyp['shear'],
+                                       perspective=self.hyp['perspective'],
+                                       border=self.mosaic_border)  # border to remove
+    return img4, labels4
+
+
+# Ancillary functions --------------------------------------------------------------------------------------------------
+def load_image(self, index):
+    # loads 1 image from dataset, returns img, original hw, resized hw
+    img = self.imgs[index]
+    if img is None:  # not cached
+        path = self.img_files[index]
+        img = cv2.imread(path)  # BGR
+        assert img is not None, 'Image Not Found ' + path
+        h0, w0 = img.shape[:2]  # orig hw
+        r = self.img_size / max(h0, w0)  # resize image to img_size
+        if r != 1:  # always resize down, only resize up if training with augmentation
+            interp = cv2.INTER_AREA if r < 1 and not self.augment else cv2.INTER_LINEAR
+            img = cv2.resize(img, (int(w0 * r), int(h0 * r)), interpolation=interp)
+        return img, (h0, w0), img.shape[:2]  # img, hw_original, hw_resized
+    else:
+        return self.imgs[index], self.img_hw0[index], self.img_hw[index]  # img, hw_original, hw_resized
+
+
+def augment_hsv(img, hgain=0.5, sgain=0.5, vgain=0.5):
+    r = np.random.uniform(-1, 1, 3) * [hgain, sgain, vgain] + 1  # random gains
+    hue, sat, val = cv2.split(cv2.cvtColor(img, cv2.COLOR_BGR2HSV))
+    dtype = img.dtype  # uint8
+
+    x = np.arange(0, 256, dtype=np.int16)
+    lut_hue = ((x * r[0]) % 180).astype(dtype)
+    lut_sat = np.clip(x * r[1], 0, 255).astype(dtype)
+    lut_val = np.clip(x * r[2], 0, 255).astype(dtype)
+
+    img_hsv = cv2.merge((cv2.LUT(hue, lut_hue), cv2.LUT(sat, lut_sat), cv2.LUT(val, lut_val))).astype(dtype)
+    cv2.cvtColor(img_hsv, cv2.COLOR_HSV2BGR, dst=img)  # no return needed
+
+    # Histogram equalization
+    # if random.random() < 0.2:
+    #     for i in range(3):
+    #         img[:, :, i] = cv2.equalizeHist(img[:, :, i])
+
+def replicate(img, labels):
+    # Replicate labels
+    h, w = img.shape[:2]
+    boxes = labels[:, 1:].astype(int)
+    x1, y1, x2, y2 = boxes.T
+    s = ((x2 - x1) + (y2 - y1)) / 2  # side length (pixels)
+    for i in s.argsort()[:round(s.size * 0.5)]:  # smallest indices
+        x1b, y1b, x2b, y2b = boxes[i]
+        bh, bw = y2b - y1b, x2b - x1b
+        yc, xc = int(random.uniform(0, h - bh)), int(random.uniform(0, w - bw))  # offset x, y
+        x1a, y1a, x2a, y2a = [xc, yc, xc + bw, yc + bh]
+        img[y1a:y2a, x1a:x2a] = img[y1b:y2b, x1b:x2b]  # img4[ymin:ymax, xmin:xmax]
+        labels = np.append(labels, [[labels[i, 0], x1a, y1a, x2a, y2a]], axis=0)
+
+    return img, labels
+
+
+def letterbox(img, new_shape=(640, 640), color=(114, 114, 114), auto=True, scaleFill=False, scaleup=True):
+    # Resize image to a 32-pixel-multiple rectangle https://github.com/ultralytics/yolov3/issues/232
+    shape = img.shape[:2]  # current shape [height, width]
+    if isinstance(new_shape, int):
+        new_shape = (new_shape, new_shape)
+
+    # Scale ratio (new / old)
+    r = min(new_shape[0] / shape[0], new_shape[1] / shape[1])
+    if not scaleup:  # only scale down, do not scale up (for better test mAP)
+        r = min(r, 1.0)
+
+    # Compute padding
+    ratio = r, r  # width, height ratios
+    new_unpad = int(round(shape[1] * r)), int(round(shape[0] * r))
+    dw, dh = new_shape[1] - new_unpad[0], new_shape[0] - new_unpad[1]  # wh padding
+    if auto:  # minimum rectangle
+        dw, dh = np.mod(dw, 64), np.mod(dh, 64)  # wh padding
+    elif scaleFill:  # stretch
+        dw, dh = 0.0, 0.0
+        new_unpad = (new_shape[1], new_shape[0])
+        ratio = new_shape[1] / shape[1], new_shape[0] / shape[0]  # width, height ratios
+
+    dw /= 2  # divide padding into 2 sides
+    dh /= 2
+
+    if shape[::-1] != new_unpad:  # resize
+        img = cv2.resize(img, new_unpad, interpolation=cv2.INTER_LINEAR)
+    top, bottom = int(round(dh - 0.1)), int(round(dh + 0.1))
+    left, right = int(round(dw - 0.1)), int(round(dw + 0.1))
+    img = cv2.copyMakeBorder(img, top, bottom, left, right, cv2.BORDER_CONSTANT, value=color)  # add border
+    return img, ratio, (dw, dh)
+
+
+def random_perspective(img, targets=(), degrees=10, translate=.1, scale=.1, shear=10, perspective=0.0, border=(0, 0)):
+    # torchvision.transforms.RandomAffine(degrees=(-10, 10), translate=(.1, .1), scale=(.9, 1.1), shear=(-10, 10))
+    # targets = [cls, xyxy]
+
+    height = img.shape[0] + border[0] * 2  # shape(h,w,c)
+    width = img.shape[1] + border[1] * 2
+
+    # Center
+    C = np.eye(3)
+    C[0, 2] = -img.shape[1] / 2  # x translation (pixels)
+    C[1, 2] = -img.shape[0] / 2  # y translation (pixels)
+
+    # Perspective
+    P = np.eye(3)
+    P[2, 0] = random.uniform(-perspective, perspective)  # x perspective (about y)
+    P[2, 1] = random.uniform(-perspective, perspective)  # y perspective (about x)
+
+    # Rotation and Scale
+    R = np.eye(3)
+    a = random.uniform(-degrees, degrees)
+    # a += random.choice([-180, -90, 0, 90])  # add 90deg rotations to small rotations
+    s = random.uniform(1 - scale, 1 + scale)
+    # s = 2 ** random.uniform(-scale, scale)
+    R[:2] = cv2.getRotationMatrix2D(angle=a, center=(0, 0), scale=s)
+
+    # Shear
+    S = np.eye(3)
+    S[0, 1] = math.tan(random.uniform(-shear, shear) * math.pi / 180)  # x shear (deg)
+    S[1, 0] = math.tan(random.uniform(-shear, shear) * math.pi / 180)  # y shear (deg)
+
+    # Translation
+    T = np.eye(3)
+    T[0, 2] = random.uniform(0.5 - translate, 0.5 + translate) * width  # x translation (pixels)
+    T[1, 2] = random.uniform(0.5 - translate, 0.5 + translate) * height  # y translation (pixels)
+
+    # Combined rotation matrix
+    M = T @ S @ R @ P @ C  # order of operations (right to left) is IMPORTANT
+    if (border[0] != 0) or (border[1] != 0) or (M != np.eye(3)).any():  # image changed
+        if perspective:
+            img = cv2.warpPerspective(img, M, dsize=(width, height), borderValue=(114, 114, 114))
+        else:  # affine
+            img = cv2.warpAffine(img, M[:2], dsize=(width, height), borderValue=(114, 114, 114))
+
+    # Visualize
+    # import matplotlib.pyplot as plt
+    # ax = plt.subplots(1, 2, figsize=(12, 6))[1].ravel()
+    # ax[0].imshow(img[:, :, ::-1])  # base
+    # ax[1].imshow(img2[:, :, ::-1])  # warped
+
+    # Transform label coordinates
+    n = len(targets)
+    if n:
+        # warp points
+        #xy = np.ones((n * 4, 3))
+        xy = np.ones((n * 8, 3))
+        xy[:, :2] = targets[:, [1, 2, 3, 4, 1, 4, 3, 2, 5, 6, 7, 8, 9, 10, 11, 12]].reshape(n * 8, 2)  # x1y1, x2y2, x1y2, x2y1
+        xy = xy @ M.T  # transform
+        if perspective:
+            xy = (xy[:, :2] / xy[:, 2:3]).reshape(n, 16)  # rescale
+        else:  # affine
+            xy = xy[:, :2].reshape(n, 16)
+
+        # create new boxes
+        x = xy[:, [0, 2, 4, 6]]
+        y = xy[:, [1, 3, 5, 7]]
+
+        landmarks = xy[:, [8, 9, 10, 11, 12, 13, 14,15]]
+        mask = np.array(targets[:, 5:] > 0, dtype=np.int32)
+        landmarks = landmarks * mask
+        landmarks = landmarks + mask - 1
+
+        landmarks = np.where(landmarks < 0, -1, landmarks)
+        landmarks[:, [0, 2, 4, 6]] = np.where(landmarks[:, [0, 2, 4, 6]] > width, -1, landmarks[:, [0, 2, 4, 6]])
+        landmarks[:, [1, 3, 5, 7]] = np.where(landmarks[:, [1, 3, 5, 7]] > height, -1,landmarks[:, [1, 3, 5, 7]])
+
+        landmarks[:, 0] = np.where(landmarks[:, 1] == -1, -1, landmarks[:, 0])
+        landmarks[:, 1] = np.where(landmarks[:, 0] == -1, -1, landmarks[:, 1])
+
+        landmarks[:, 2] = np.where(landmarks[:, 3] == -1, -1, landmarks[:, 2])
+        landmarks[:, 3] = np.where(landmarks[:, 2] == -1, -1, landmarks[:, 3])
+
+        landmarks[:, 4] = np.where(landmarks[:, 5] == -1, -1, landmarks[:, 4])
+        landmarks[:, 5] = np.where(landmarks[:, 4] == -1, -1, landmarks[:, 5])
+
+        landmarks[:, 6] = np.where(landmarks[:, 7] == -1, -1, landmarks[:, 6])
+        landmarks[:, 7] = np.where(landmarks[:, 6] == -1, -1, landmarks[:, 7])
+
+        # landmarks[:, 8] = np.where(landmarks[:, 9] == -1, -1, landmarks[:, 8])
+        # landmarks[:, 9] = np.where(landmarks[:, 8] == -1, -1, landmarks[:, 9])
+
+        targets[:,5:] = landmarks
+
+        xy = np.concatenate((x.min(1), y.min(1), x.max(1), y.max(1))).reshape(4, n).T
+
+        # # apply angle-based reduction of bounding boxes
+        # radians = a * math.pi / 180
+        # reduction = max(abs(math.sin(radians)), abs(math.cos(radians))) ** 0.5
+        # x = (xy[:, 2] + xy[:, 0]) / 2
+        # y = (xy[:, 3] + xy[:, 1]) / 2
+        # w = (xy[:, 2] - xy[:, 0]) * reduction
+        # h = (xy[:, 3] - xy[:, 1]) * reduction
+        # xy = np.concatenate((x - w / 2, y - h / 2, x + w / 2, y + h / 2)).reshape(4, n).T
+
+        # clip boxes
+        xy[:, [0, 2]] = xy[:, [0, 2]].clip(0, width)
+        xy[:, [1, 3]] = xy[:, [1, 3]].clip(0, height)
+
+        # filter candidates
+        i = box_candidates(box1=targets[:, 1:5].T * s, box2=xy.T)
+        targets = targets[i]
+        targets[:, 1:5] = xy[i]
+
+    return img, targets
+
+
+def box_candidates(box1, box2, wh_thr=2, ar_thr=20, area_thr=0.1):  # box1(4,n), box2(4,n)
+    # Compute candidate boxes: box1 before augment, box2 after augment, wh_thr (pixels), aspect_ratio_thr, area_ratio
+    w1, h1 = box1[2] - box1[0], box1[3] - box1[1]
+    w2, h2 = box2[2] - box2[0], box2[3] - box2[1]
+    ar = np.maximum(w2 / (h2 + 1e-16), h2 / (w2 + 1e-16))  # aspect ratio
+    return (w2 > wh_thr) & (h2 > wh_thr) & (w2 * h2 / (w1 * h1 + 1e-16) > area_thr) & (ar < ar_thr)  # candidates
+
+
+def cutout(image, labels):
+    # Applies image cutout augmentation https://arxiv.org/abs/1708.04552
+    h, w = image.shape[:2]
+
+    def bbox_ioa(box1, box2):
+        # Returns the intersection over box2 area given box1, box2. box1 is 4, box2 is nx4. boxes are x1y1x2y2
+        box2 = box2.transpose()
+
+        # Get the coordinates of bounding boxes
+        b1_x1, b1_y1, b1_x2, b1_y2 = box1[0], box1[1], box1[2], box1[3]
+        b2_x1, b2_y1, b2_x2, b2_y2 = box2[0], box2[1], box2[2], box2[3]
+
+        # Intersection area
+        inter_area = (np.minimum(b1_x2, b2_x2) - np.maximum(b1_x1, b2_x1)).clip(0) * \
+                     (np.minimum(b1_y2, b2_y2) - np.maximum(b1_y1, b2_y1)).clip(0)
+
+        # box2 area
+        box2_area = (b2_x2 - b2_x1) * (b2_y2 - b2_y1) + 1e-16
+
+        # Intersection over box2 area
+        return inter_area / box2_area
+
+    # create random masks
+    scales = [0.5] * 1 + [0.25] * 2 + [0.125] * 4 + [0.0625] * 8 + [0.03125] * 16  # image size fraction
+    for s in scales:
+        mask_h = random.randint(1, int(h * s))
+        mask_w = random.randint(1, int(w * s))
+
+        # box
+        xmin = max(0, random.randint(0, w) - mask_w // 2)
+        ymin = max(0, random.randint(0, h) - mask_h // 2)
+        xmax = min(w, xmin + mask_w)
+        ymax = min(h, ymin + mask_h)
+
+        # apply random color mask
+        image[ymin:ymax, xmin:xmax] = [random.randint(64, 191) for _ in range(3)]
+
+        # return unobscured labels
+        if len(labels) and s > 0.03:
+            box = np.array([xmin, ymin, xmax, ymax], dtype=np.float32)
+            ioa = bbox_ioa(box, labels[:, 1:5])  # intersection over area
+            labels = labels[ioa < 0.60]  # remove >60% obscured labels
+
+    return labels
+
+
+def create_folder(path='./new'):
+    # Create folder
+    if os.path.exists(path):
+        shutil.rmtree(path)  # delete output folder
+    os.makedirs(path)  # make new output folder
+
+
+def flatten_recursive(path='../coco128'):
+    # Flatten a recursive directory by bringing all files to top level
+    new_path = Path(path + '_flat')
+    create_folder(new_path)
+    for file in tqdm(glob.glob(str(Path(path)) + '/**/*.*', recursive=True)):
+        shutil.copyfile(file, new_path / Path(file).name)
+
+
+def extract_boxes(path='../coco128/'):  # from utils.datasets import *; extract_boxes('../coco128')
+    # Convert detection dataset into classification dataset, with one directory per class
+
+    path = Path(path)  # images dir
+    shutil.rmtree(path / 'classifier') if (path / 'classifier').is_dir() else None  # remove existing
+    files = list(path.rglob('*.*'))
+    n = len(files)  # number of files
+    for im_file in tqdm(files, total=n):
+        if im_file.suffix[1:] in img_formats:
+            # image
+            im = cv2.imread(str(im_file))[..., ::-1]  # BGR to RGB
+            h, w = im.shape[:2]
+
+            # labels
+            lb_file = Path(img2label_paths([str(im_file)])[0])
+            if Path(lb_file).exists():
+                with open(lb_file, 'r') as f:
+                    lb = np.array([x.split() for x in f.read().strip().splitlines()], dtype=np.float32)  # labels
+
+                for j, x in enumerate(lb):
+                    c = int(x[0])  # class
+                    f = (path / 'classifier') / f'{c}' / f'{path.stem}_{im_file.stem}_{j}.jpg'  # new filename
+                    if not f.parent.is_dir():
+                        f.parent.mkdir(parents=True)
+
+                    b = x[1:] * [w, h, w, h]  # box
+                    # b[2:] = b[2:].max()  # rectangle to square
+                    b[2:] = b[2:] * 1.2 + 3  # pad
+                    b = xywh2xyxy(b.reshape(-1, 4)).ravel().astype(np.int)
+
+                    b[[0, 2]] = np.clip(b[[0, 2]], 0, w)  # clip boxes outside of image
+                    b[[1, 3]] = np.clip(b[[1, 3]], 0, h)
+                    assert cv2.imwrite(str(f), im[b[1]:b[3], b[0]:b[2]]), f'box failure in {f}'
+
+
+def autosplit(path='../coco128', weights=(0.9, 0.1, 0.0)):  # from utils.datasets import *; autosplit('../coco128')
+    """ Autosplit a dataset into train/val/test splits and save path/autosplit_*.txt files
+    # Arguments
+        path:       Path to images directory
+        weights:    Train, val, test weights (list)
+    """
+    path = Path(path)  # images dir
+    files = list(path.rglob('*.*'))
+    n = len(files)  # number of files
+    indices = random.choices([0, 1, 2], weights=weights, k=n)  # assign each image to a split
+    txt = ['autosplit_train.txt', 'autosplit_val.txt', 'autosplit_test.txt']  # 3 txt files
+    [(path / x).unlink() for x in txt if (path / x).exists()]  # remove existing
+    for i, img in tqdm(zip(indices, files), total=n):
+        if img.suffix[1:] in img_formats:
+            with open(path / txt[i], 'a') as f:
+                f.write(str(img) + '\n')  # add image to txt file

utils/general.py

@@ -0,0 +1,647 @@
+# General utils
+
+import glob
+import logging
+import math
+import os
+import random
+import re
+import subprocess
+import time
+from pathlib import Path
+
+import cv2
+import numpy as np
+import torch
+import torchvision
+import yaml
+
+from utils.google_utils import gsutil_getsize
+from utils.metrics import fitness
+from utils.torch_utils import init_torch_seeds
+
+# Settings
+torch.set_printoptions(linewidth=320, precision=5, profile='long')
+np.set_printoptions(linewidth=320, formatter={'float_kind': '{:11.5g}'.format})  # format short g, %precision=5
+cv2.setNumThreads(0)  # prevent OpenCV from multithreading (incompatible with PyTorch DataLoader)
+os.environ['NUMEXPR_MAX_THREADS'] = str(min(os.cpu_count(), 8))  # NumExpr max threads
+
+
+def set_logging(rank=-1):
+    logging.basicConfig(
+        format="%(message)s",
+        level=logging.INFO if rank in [-1, 0] else logging.WARN)
+
+
+def init_seeds(seed=0):
+    # Initialize random number generator (RNG) seeds
+    random.seed(seed)
+    np.random.seed(seed)
+    init_torch_seeds(seed)
+
+
+def get_latest_run(search_dir='.'):
+    # Return path to most recent 'last.pt' in /runs (i.e. to --resume from)
+    last_list = glob.glob(f'{search_dir}/**/last*.pt', recursive=True)
+    return max(last_list, key=os.path.getctime) if last_list else ''
+
+
+def check_online():
+    # Check internet connectivity
+    import socket
+    try:
+        socket.create_connection(("1.1.1.1", 53))  # check host accesability
+        return True
+    except OSError:
+        return False
+
+
+def check_git_status():
+    # Recommend 'git pull' if code is out of date
+    print(colorstr('github: '), end='')
+    try:
+        assert Path('.git').exists(), 'skipping check (not a git repository)'
+        assert not Path('/workspace').exists(), 'skipping check (Docker image)'  # not Path('/.dockerenv').exists()
+        assert check_online(), 'skipping check (offline)'
+
+        cmd = 'git fetch && git config --get remote.origin.url'  # github repo url
+        url = subprocess.check_output(cmd, shell=True).decode()[:-1]
+        cmd = 'git rev-list $(git rev-parse --abbrev-ref HEAD)..origin/master --count'  # commits behind
+        n = int(subprocess.check_output(cmd, shell=True))
+        if n > 0:
+            print(f"⚠️ WARNING: code is out of date by {n} {'commits' if n > 1 else 'commmit'}. "
+                  f"Use 'git pull' to update or 'git clone {url}' to download latest.")
+        else:
+            print(f'up to date with {url} ✅')
+    except Exception as e:
+        print(e)
+
+
+def check_requirements(file='requirements.txt'):
+    # Check installed dependencies meet requirements
+    import pkg_resources
+    requirements = pkg_resources.parse_requirements(Path(file).open())
+    requirements = [x.name + ''.join(*x.specs) if len(x.specs) else x.name for x in requirements]
+    pkg_resources.require(requirements)  # DistributionNotFound or VersionConflict exception if requirements not met
+
+
+def check_img_size(img_size, s=32):
+    # Verify img_size is a multiple of stride s
+    new_size = make_divisible(img_size, int(s))  # ceil gs-multiple
+    if new_size != img_size:
+        print('WARNING: --img-size %g must be multiple of max stride %g, updating to %g' % (img_size, s, new_size))
+    return new_size
+
+
+def check_file(file):
+    # Search for file if not found
+    if os.path.isfile(file) or file == '':
+        return file
+    else:
+        files = glob.glob('./**/' + file, recursive=True)  # find file
+        assert len(files), 'File Not Found: %s' % file  # assert file was found
+        assert len(files) == 1, "Multiple files match '%s', specify exact path: %s" % (file, files)  # assert unique
+        return files[0]  # return file
+
+
+def check_dataset(dict):
+    # Download dataset if not found locally
+    val, s = dict.get('val'), dict.get('download')
+    if val and len(val):
+        val = [Path(x).resolve() for x in (val if isinstance(val, list) else [val])]  # val path
+        if not all(x.exists() for x in val):
+            print('\nWARNING: Dataset not found, nonexistent paths: %s' % [str(x) for x in val if not x.exists()])
+            if s and len(s):  # download script
+                print('Downloading %s ...' % s)
+                if s.startswith('http') and s.endswith('.zip'):  # URL
+                    f = Path(s).name  # filename
+                    torch.hub.download_url_to_file(s, f)
+                    r = os.system('unzip -q %s -d ../ && rm %s' % (f, f))  # unzip
+                else:  # bash script
+                    r = os.system(s)
+                print('Dataset autodownload %s\n' % ('success' if r == 0 else 'failure'))  # analyze return value
+            else:
+                raise Exception('Dataset not found.')
+
+
+def make_divisible(x, divisor):
+    # Returns x evenly divisible by divisor
+    return math.ceil(x / divisor) * divisor
+
+
+def clean_str(s):
+    # Cleans a string by replacing special characters with underscore _
+    return re.sub(pattern="[|@#!¡·$€%&()=?¿^*;:,¨´><+]", repl="_", string=s)
+
+
+def one_cycle(y1=0.0, y2=1.0, steps=100):
+    # lambda function for sinusoidal ramp from y1 to y2
+    return lambda x: ((1 - math.cos(x * math.pi / steps)) / 2) * (y2 - y1) + y1
+
+
+def colorstr(*input):
+    # Colors a string https://en.wikipedia.org/wiki/ANSI_escape_code, i.e.  colorstr('blue', 'hello world')
+    *args, string = input if len(input) > 1 else ('blue', 'bold', input[0])  # color arguments, string
+    colors = {'black': '\033[30m',  # basic colors
+              'red': '\033[31m',
+              'green': '\033[32m',
+              'yellow': '\033[33m',
+              'blue': '\033[34m',
+              'magenta': '\033[35m',
+              'cyan': '\033[36m',
+              'white': '\033[37m',
+              'bright_black': '\033[90m',  # bright colors
+              'bright_red': '\033[91m',
+              'bright_green': '\033[92m',
+              'bright_yellow': '\033[93m',
+              'bright_blue': '\033[94m',
+              'bright_magenta': '\033[95m',
+              'bright_cyan': '\033[96m',
+              'bright_white': '\033[97m',
+              'end': '\033[0m',  # misc
+              'bold': '\033[1m',
+              'underline': '\033[4m'}
+    return ''.join(colors[x] for x in args) + f'{string}' + colors['end']
+
+
+def labels_to_class_weights(labels, nc=80):
+    # Get class weights (inverse frequency) from training labels
+    if labels[0] is None:  # no labels loaded
+        return torch.Tensor()
+
+    labels = np.concatenate(labels, 0)  # labels.shape = (866643, 5) for COCO
+    classes = labels[:, 0].astype(np.int)  # labels = [class xywh]
+    weights = np.bincount(classes, minlength=nc)  # occurrences per class
+
+    # Prepend gridpoint count (for uCE training)
+    # gpi = ((320 / 32 * np.array([1, 2, 4])) ** 2 * 3).sum()  # gridpoints per image
+    # weights = np.hstack([gpi * len(labels)  - weights.sum() * 9, weights * 9]) ** 0.5  # prepend gridpoints to start
+
+    weights[weights == 0] = 1  # replace empty bins with 1
+    weights = 1 / weights  # number of targets per class
+    weights /= weights.sum()  # normalize
+    return torch.from_numpy(weights)
+
+
+def labels_to_image_weights(labels, nc=80, class_weights=np.ones(80)):
+    # Produces image weights based on class_weights and image contents
+    class_counts = np.array([np.bincount(x[:, 0].astype(np.int), minlength=nc) for x in labels])
+    image_weights = (class_weights.reshape(1, nc) * class_counts).sum(1)
+    # index = random.choices(range(n), weights=image_weights, k=1)  # weight image sample
+    return image_weights
+
+
+def coco80_to_coco91_class():  # converts 80-index (val2014) to 91-index (paper)
+    # https://tech.amikelive.com/node-718/what-object-categories-labels-are-in-coco-dataset/
+    # a = np.loadtxt('data/coco.names', dtype='str', delimiter='\n')
+    # b = np.loadtxt('data/coco_paper.names', dtype='str', delimiter='\n')
+    # x1 = [list(a[i] == b).index(True) + 1 for i in range(80)]  # darknet to coco
+    # x2 = [list(b[i] == a).index(True) if any(b[i] == a) else None for i in range(91)]  # coco to darknet
+    x = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 27, 28, 31, 32, 33, 34,
+         35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63,
+         64, 65, 67, 70, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 84, 85, 86, 87, 88, 89, 90]
+    return x
+
+
+def xyxy2xywh(x):
+    # Convert nx4 boxes from [x1, y1, x2, y2] to [x, y, w, h] where xy1=top-left, xy2=bottom-right
+    y = x.clone() if isinstance(x, torch.Tensor) else np.copy(x)
+    y[:, 0] = (x[:, 0] + x[:, 2]) / 2  # x center
+    y[:, 1] = (x[:, 1] + x[:, 3]) / 2  # y center
+    y[:, 2] = x[:, 2] - x[:, 0]  # width
+    y[:, 3] = x[:, 3] - x[:, 1]  # height
+    return y
+
+
+def xywh2xyxy(x):
+    # Convert nx4 boxes from [x, y, w, h] to [x1, y1, x2, y2] where xy1=top-left, xy2=bottom-right
+    y = x.clone() if isinstance(x, torch.Tensor) else np.copy(x)
+    y[:, 0] = x[:, 0] - x[:, 2] / 2  # top left x
+    y[:, 1] = x[:, 1] - x[:, 3] / 2  # top left y
+    y[:, 2] = x[:, 0] + x[:, 2] / 2  # bottom right x
+    y[:, 3] = x[:, 1] + x[:, 3] / 2  # bottom right y
+    return y
+
+
+def xywhn2xyxy(x, w=640, h=640, padw=32, padh=32):
+    # Convert nx4 boxes from [x, y, w, h] normalized to [x1, y1, x2, y2] where xy1=top-left, xy2=bottom-right
+    y = x.clone() if isinstance(x, torch.Tensor) else np.copy(x)
+    y[:, 0] = w * (x[:, 0] - x[:, 2] / 2) + padw  # top left x
+    y[:, 1] = h * (x[:, 1] - x[:, 3] / 2) + padh  # top left y
+    y[:, 2] = w * (x[:, 0] + x[:, 2] / 2) + padw  # bottom right x
+    y[:, 3] = h * (x[:, 1] + x[:, 3] / 2) + padh  # bottom right y
+    return y
+
+
+def scale_coords(img1_shape, coords, img0_shape, ratio_pad=None):
+    # Rescale coords (xyxy) from img1_shape to img0_shape
+    if ratio_pad is None:  # calculate from img0_shape
+        gain = min(img1_shape[0] / img0_shape[0], img1_shape[1] / img0_shape[1])  # gain  = old / new
+        pad = (img1_shape[1] - img0_shape[1] * gain) / 2, (img1_shape[0] - img0_shape[0] * gain) / 2  # wh padding
+    else:
+        gain = ratio_pad[0][0]
+        pad = ratio_pad[1]
+
+    coords[:, [0, 2]] -= pad[0]  # x padding
+    coords[:, [1, 3]] -= pad[1]  # y padding
+    coords[:, :4] /= gain
+    clip_coords(coords, img0_shape)
+    return coords
+
+
+def clip_coords(boxes, img_shape):
+    # Clip bounding xyxy bounding boxes to image shape (height, width)
+    boxes[:, 0].clamp_(0, img_shape[1])  # x1
+    boxes[:, 1].clamp_(0, img_shape[0])  # y1
+    boxes[:, 2].clamp_(0, img_shape[1])  # x2
+    boxes[:, 3].clamp_(0, img_shape[0])  # y2
+
+
+def bbox_iou(box1, box2, x1y1x2y2=True, GIoU=False, DIoU=False, CIoU=False, eps=1e-9):
+    # Returns the IoU of box1 to box2. box1 is 4, box2 is nx4
+    box2 = box2.T
+
+    # Get the coordinates of bounding boxes
+    if x1y1x2y2:  # x1, y1, x2, y2 = box1
+        b1_x1, b1_y1, b1_x2, b1_y2 = box1[0], box1[1], box1[2], box1[3]
+        b2_x1, b2_y1, b2_x2, b2_y2 = box2[0], box2[1], box2[2], box2[3]
+    else:  # transform from xywh to xyxy
+        b1_x1, b1_x2 = box1[0] - box1[2] / 2, box1[0] + box1[2] / 2
+        b1_y1, b1_y2 = box1[1] - box1[3] / 2, box1[1] + box1[3] / 2
+        b2_x1, b2_x2 = box2[0] - box2[2] / 2, box2[0] + box2[2] / 2
+        b2_y1, b2_y2 = box2[1] - box2[3] / 2, box2[1] + box2[3] / 2
+
+    # Intersection area
+    inter = (torch.min(b1_x2, b2_x2) - torch.max(b1_x1, b2_x1)).clamp(0) * \
+            (torch.min(b1_y2, b2_y2) - torch.max(b1_y1, b2_y1)).clamp(0)
+
+    # Union Area
+    w1, h1 = b1_x2 - b1_x1, b1_y2 - b1_y1 + eps
+    w2, h2 = b2_x2 - b2_x1, b2_y2 - b2_y1 + eps
+    union = w1 * h1 + w2 * h2 - inter + eps
+
+    iou = inter / union
+    if GIoU or DIoU or CIoU:
+        # convex (smallest enclosing box) width
+        cw = torch.max(b1_x2, b2_x2) - torch.min(b1_x1, b2_x1)
+        ch = torch.max(b1_y2, b2_y2) - torch.min(b1_y1, b2_y1)  # convex height
+        if CIoU or DIoU:  # Distance or Complete IoU https://arxiv.org/abs/1911.08287v1
+            c2 = cw ** 2 + ch ** 2 + eps  # convex diagonal squared
+            rho2 = ((b2_x1 + b2_x2 - b1_x1 - b1_x2) ** 2 +
+                    (b2_y1 + b2_y2 - b1_y1 - b1_y2) ** 2) / 4  # center distance squared
+            if DIoU:
+                return iou - rho2 / c2  # DIoU
+            elif CIoU:  # https://github.com/Zzh-tju/DIoU-SSD-pytorch/blob/master/utils/box/box_utils.py#L47
+                v = (4 / math.pi ** 2) * \
+                    torch.pow(torch.atan(w2 / h2) - torch.atan(w1 / h1), 2)
+                with torch.no_grad():
+                    alpha = v / ((1 + eps) - iou + v)
+                return iou - (rho2 / c2 + v * alpha)  # CIoU
+        else:  # GIoU https://arxiv.org/pdf/1902.09630.pdf
+            c_area = cw * ch + eps  # convex area
+            return iou - (c_area - union) / c_area  # GIoU
+    else:
+        return iou  # IoU
+
+
+def box_iou(box1, box2):
+    # https://github.com/pytorch/vision/blob/master/torchvision/ops/boxes.py
+    """
+    Return intersection-over-union (Jaccard index) of boxes.
+    Both sets of boxes are expected to be in (x1, y1, x2, y2) format.
+    Arguments:
+        box1 (Tensor[N, 4])
+        box2 (Tensor[M, 4])
+    Returns:
+        iou (Tensor[N, M]): the NxM matrix containing the pairwise
+            IoU values for every element in boxes1 and boxes2
+    """
+
+    def box_area(box):
+        # box = 4xn
+        return (box[2] - box[0]) * (box[3] - box[1])
+
+    area1 = box_area(box1.T)
+    area2 = box_area(box2.T)
+
+    # inter(N,M) = (rb(N,M,2) - lt(N,M,2)).clamp(0).prod(2)
+    inter = (torch.min(box1[:, None, 2:], box2[:, 2:]) -
+             torch.max(box1[:, None, :2], box2[:, :2])).clamp(0).prod(2)
+    # iou = inter / (area1 + area2 - inter)
+    return inter / (area1[:, None] + area2 - inter)
+
+
+def wh_iou(wh1, wh2):
+    # Returns the nxm IoU matrix. wh1 is nx2, wh2 is mx2
+    wh1 = wh1[:, None]  # [N,1,2]
+    wh2 = wh2[None]  # [1,M,2]
+    inter = torch.min(wh1, wh2).prod(2)  # [N,M]
+    # iou = inter / (area1 + area2 - inter)
+    return inter / (wh1.prod(2) + wh2.prod(2) - inter)
+
+def jaccard_diou(box_a, box_b, iscrowd:bool=False):
+    use_batch = True
+    if box_a.dim() == 2:
+        use_batch = False
+        box_a = box_a[None, ...]
+        box_b = box_b[None, ...]
+
+    inter = intersect(box_a, box_b)
+    area_a = ((box_a[:, :, 2]-box_a[:, :, 0]) *
+              (box_a[:, :, 3]-box_a[:, :, 1])).unsqueeze(2).expand_as(inter)  # [A,B]
+    area_b = ((box_b[:, :, 2]-box_b[:, :, 0]) *
+              (box_b[:, :, 3]-box_b[:, :, 1])).unsqueeze(1).expand_as(inter)  # [A,B]
+    union = area_a + area_b - inter
+    x1 = ((box_a[:, :, 2]+box_a[:, :, 0]) / 2).unsqueeze(2).expand_as(inter)
+    y1 = ((box_a[:, :, 3]+box_a[:, :, 1]) / 2).unsqueeze(2).expand_as(inter)
+    x2 = ((box_b[:, :, 2]+box_b[:, :, 0]) / 2).unsqueeze(1).expand_as(inter)
+    y2 = ((box_b[:, :, 3]+box_b[:, :, 1]) / 2).unsqueeze(1).expand_as(inter)
+
+    t1 = box_a[:, :, 1].unsqueeze(2).expand_as(inter)
+    b1 = box_a[:, :, 3].unsqueeze(2).expand_as(inter)
+    l1 = box_a[:, :, 0].unsqueeze(2).expand_as(inter)
+    r1 = box_a[:, :, 2].unsqueeze(2).expand_as(inter)
+
+    t2 = box_b[:, :, 1].unsqueeze(1).expand_as(inter)
+    b2 = box_b[:, :, 3].unsqueeze(1).expand_as(inter)
+    l2 = box_b[:, :, 0].unsqueeze(1).expand_as(inter)
+    r2 = box_b[:, :, 2].unsqueeze(1).expand_as(inter)
+
+    cr = torch.max(r1, r2)
+    cl = torch.min(l1, l2)
+    ct = torch.min(t1, t2)
+    cb = torch.max(b1, b2)
+    D = (((x2 - x1)**2 + (y2 - y1)**2) / ((cr-cl)**2 + (cb-ct)**2 + 1e-7))
+    out = inter / area_a if iscrowd else inter / (union + 1e-7) - D ** 0.7
+    return out if use_batch else out.squeeze(0)
+
+
+def non_max_suppression_face(prediction, conf_thres=0.25, iou_thres=0.45, classes=None, agnostic=False, labels=()):
+    """Performs Non-Maximum Suppression (NMS) on inference results
+    Returns:
+         detections with shape: nx6 (x1, y1, x2, y2, conf, cls)
+    """
+
+    nc = prediction.shape[2] - 13  # number of classes
+    xc = prediction[..., 4] > conf_thres  # candidates
+
+    # Settings
+    min_wh, max_wh = 2, 4096  # (pixels) minimum and maximum box width and height
+    time_limit = 10.0  # seconds to quit after
+    redundant = True  # require redundant detections
+    multi_label = nc > 1  # multiple labels per box (adds 0.5ms/img)
+    multi_label=False
+    merge = False  # use merge-NMS
+
+    t = time.time()
+    output = [torch.zeros((0, 14), device=prediction.device)] * prediction.shape[0]
+    for xi, x in enumerate(prediction):  # image index, image inference
+        # Apply constraints
+        # x[((x[..., 2:4] < min_wh) | (x[..., 2:4] > max_wh)).any(1), 4] = 0  # width-height
+        x = x[xc[xi]]  # confidence
+
+        # Cat apriori labels if autolabelling
+        if labels and len(labels[xi]):
+            l = labels[xi]
+            v = torch.zeros((len(l), nc + 13), device=x.device)
+            v[:, :4] = l[:, 1:5]  # box
+            v[:, 4] = 1.0  # conf
+            v[range(len(l)), l[:, 0].long() + 13] = 1.0  # cls
+            x = torch.cat((x, v), 0)
+
+        # If none remain process next image
+        if not x.shape[0]:
+            continue
+
+        # Compute conf
+        x[:, 13:] *= x[:, 4:5]  # conf = obj_conf * cls_conf
+
+        # Box (center x, center y, width, height) to (x1, y1, x2, y2)
+        box = xywh2xyxy(x[:, :4])
+
+        # Detections matrix nx6 (xyxy, conf, landmarks, cls)
+        if multi_label:
+            i, j = (x[:, 13:] > conf_thres).nonzero(as_tuple=False).T
+            x = torch.cat((box[i], x[i, j + 13, None], x[i, 5:13] ,j[:, None].float()), 1)
+        else:  # best class only
+            conf, j = x[:, 13:].max(1, keepdim=True)
+            x = torch.cat((box, conf, x[:, 5:13], j.float()), 1)[conf.view(-1) > conf_thres]
+
+        # Filter by class
+        if classes is not None:
+            x = x[(x[:, 5:6] == torch.tensor(classes, device=x.device)).any(1)]
+
+        # If none remain process next image
+        n = x.shape[0]  # number of boxes
+        if not n:
+            continue
+
+        # Batched NMS
+        c = x[:, 13:14] * (0 if agnostic else max_wh)  # classes
+        boxes, scores = x[:, :4] + c, x[:, 4]  # boxes (offset by class), scores
+        i = torchvision.ops.nms(boxes, scores, iou_thres)  # NMS
+        #if i.shape[0] > max_det:  # limit detections
+        #    i = i[:max_det]
+        if merge and (1 < n < 3E3):  # Merge NMS (boxes merged using weighted mean)
+            # update boxes as boxes(i,4) = weights(i,n) * boxes(n,4)
+            iou = box_iou(boxes[i], boxes) > iou_thres  # iou matrix
+            weights = iou * scores[None]  # box weights
+            x[i, :4] = torch.mm(weights, x[:, :4]).float() / weights.sum(1, keepdim=True)  # merged boxes
+            if redundant:
+                i = i[iou.sum(1) > 1]  # require redundancy
+
+        output[xi] = x[i]
+        if (time.time() - t) > time_limit:
+            break  # time limit exceeded
+
+    return output
+
+
+def non_max_suppression(prediction, conf_thres=0.25, iou_thres=0.45, classes=None, agnostic=False, labels=()):
+    """Performs Non-Maximum Suppression (NMS) on inference results
+
+    Returns:
+         detections with shape: nx6 (x1, y1, x2, y2, conf, cls)
+    """
+
+    nc = prediction.shape[2] - 5  # number of classes
+    xc = prediction[..., 4] > conf_thres  # candidates
+
+    # Settings
+    # (pixels) minimum and maximum box width and height
+    min_wh, max_wh = 2, 4096
+    #max_det = 300  # maximum number of detections per image
+    #max_nms = 30000  # maximum number of boxes into torchvision.ops.nms()
+    time_limit = 10.0  # seconds to quit after
+    redundant = True  # require redundant detections
+    multi_label = nc > 1  # multiple labels per box (adds 0.5ms/img)
+    merge = False  # use merge-NMS
+
+    t = time.time()
+    output = [torch.zeros((0, 6), device=prediction.device)] * prediction.shape[0]
+    for xi, x in enumerate(prediction):  # image index, image inference
+        # Apply constraints
+        # x[((x[..., 2:4] < min_wh) | (x[..., 2:4] > max_wh)).any(1), 4] = 0  # width-height
+        x = x[xc[xi]]  # confidence
+
+        # Cat apriori labels if autolabelling
+        if labels and len(labels[xi]):
+            l = labels[xi]
+            v = torch.zeros((len(l), nc + 5), device=x.device)
+            v[:, :4] = l[:, 1:5]  # box
+            v[:, 4] = 1.0  # conf
+            v[range(len(l)), l[:, 0].long() + 5] = 1.0  # cls
+            x = torch.cat((x, v), 0)
+
+        # If none remain process next image
+        if not x.shape[0]:
+            continue
+
+        # Compute conf
+        x[:, 5:] *= x[:, 4:5]  # conf = obj_conf * cls_conf
+
+        # Box (center x, center y, width, height) to (x1, y1, x2, y2)
+        box = xywh2xyxy(x[:, :4])
+
+        # Detections matrix nx6 (xyxy, conf, cls)
+        if multi_label:
+            i, j = (x[:, 5:] > conf_thres).nonzero(as_tuple=False).T
+            x = torch.cat((box[i], x[i, j + 5, None], j[:, None].float()), 1)
+        else:  # best class only
+            conf, j = x[:, 5:].max(1, keepdim=True)
+            x = torch.cat((box, conf, j.float()), 1)[
+                conf.view(-1) > conf_thres]
+
+        # Filter by class
+        if classes is not None:
+            x = x[(x[:, 5:6] == torch.tensor(classes, device=x.device)).any(1)]
+
+        # Apply finite constraint
+        # if not torch.isfinite(x).all():
+        #     x = x[torch.isfinite(x).all(1)]
+
+        # Check shape
+        n = x.shape[0]  # number of boxes
+        if not n:  # no boxes
+            continue
+        #elif n > max_nms:  # excess boxes
+        #    x = x[x[:, 4].argsort(descending=True)[:max_nms]]  # sort by confidence
+        x = x[x[:, 4].argsort(descending=True)]  # sort by confidence
+
+        # Batched NMS
+        c = x[:, 5:6] * (0 if agnostic else max_wh)  # classes
+        boxes, scores = x[:, :4] + c, x[:, 4]  # boxes (offset by class), scores
+        i = torchvision.ops.nms(boxes, scores, iou_thres)  # NMS
+        #if i.shape[0] > max_det:  # limit detections
+        #    i = i[:max_det]
+        if merge and (1 < n < 3E3):  # Merge NMS (boxes merged using weighted mean)
+            # update boxes as boxes(i,4) = weights(i,n) * boxes(n,4)
+            iou = box_iou(boxes[i], boxes) > iou_thres  # iou matrix
+            weights = iou * scores[None]  # box weights
+            x[i, :4] = torch.mm(weights, x[:, :4]).float() / weights.sum(1, keepdim=True)  # merged boxes
+            if redundant:
+                i = i[iou.sum(1) > 1]  # require redundancy
+
+        output[xi] = x[i]
+        if (time.time() - t) > time_limit:
+            print(f'WARNING: NMS time limit {time_limit}s exceeded')
+            break  # time limit exceeded
+
+    return output
+
+
+def strip_optimizer(f='weights/best.pt', s=''):  # from utils.general import *; strip_optimizer()
+    # Strip optimizer from 'f' to finalize training, optionally save as 's'
+    x = torch.load(f, map_location=torch.device('cpu'))
+    for key in 'optimizer', 'training_results', 'wandb_id':
+        x[key] = None
+    x['epoch'] = -1
+    x['model'].half()  # to FP16
+    for p in x['model'].parameters():
+        p.requires_grad = False
+    torch.save(x, s or f)
+    mb = os.path.getsize(s or f) / 1E6  # filesize
+    print('Optimizer stripped from %s,%s %.1fMB' % (f, (' saved as %s,' % s) if s else '', mb))
+
+
+def print_mutation(hyp, results, yaml_file='hyp_evolved.yaml', bucket=''):
+    # Print mutation results to evolve.txt (for use with train.py --evolve)
+    a = '%10s' * len(hyp) % tuple(hyp.keys())  # hyperparam keys
+    b = '%10.3g' * len(hyp) % tuple(hyp.values())  # hyperparam values
+    c = '%10.4g' * len(results) % results  # results (P, R, mAP@0.5, mAP@0.5:0.95, val_losses x 3)
+    print('\n%s\n%s\nEvolved fitness: %s\n' % (a, b, c))
+
+    if bucket:
+        url = 'gs://%s/evolve.txt' % bucket
+        if gsutil_getsize(url) > (os.path.getsize('evolve.txt') if os.path.exists('evolve.txt') else 0):
+            os.system('gsutil cp %s .' % url)  # download evolve.txt if larger than local
+
+    with open('evolve.txt', 'a') as f:  # append result
+        f.write(c + b + '\n')
+    x = np.unique(np.loadtxt('evolve.txt', ndmin=2), axis=0)  # load unique rows
+    x = x[np.argsort(-fitness(x))]  # sort
+    np.savetxt('evolve.txt', x, '%10.3g')  # save sort by fitness
+
+    # Save yaml
+    for i, k in enumerate(hyp.keys()):
+        hyp[k] = float(x[0, i + 7])
+    with open(yaml_file, 'w') as f:
+        results = tuple(x[0, :7])
+        c = '%10.4g' * len(results) % results  # results (P, R, mAP@0.5, mAP@0.5:0.95, val_losses x 3)
+        f.write('# Hyperparameter Evolution Results\n# Generations: %g\n# Metrics: ' % len(x) + c + '\n\n')
+        yaml.dump(hyp, f, sort_keys=False)
+
+    if bucket:
+        os.system('gsutil cp evolve.txt %s gs://%s' % (yaml_file, bucket))  # upload
+
+
+def apply_classifier(x, model, img, im0):
+    # applies a second stage classifier to yolo outputs
+    im0 = [im0] if isinstance(im0, np.ndarray) else im0
+    for i, d in enumerate(x):  # per image
+        if d is not None and len(d):
+            d = d.clone()
+
+            # Reshape and pad cutouts
+            b = xyxy2xywh(d[:, :4])  # boxes
+            b[:, 2:] = b[:, 2:].max(1)[0].unsqueeze(1)  # rectangle to square
+            b[:, 2:] = b[:, 2:] * 1.3 + 30  # pad
+            d[:, :4] = xywh2xyxy(b).long()
+
+            # Rescale boxes from img_size to im0 size
+            scale_coords(img.shape[2:], d[:, :4], im0[i].shape)
+
+            # Classes
+            pred_cls1 = d[:, 5].long()
+            ims = []
+            for j, a in enumerate(d):  # per item
+                cutout = im0[i][int(a[1]):int(a[3]), int(a[0]):int(a[2])]
+                im = cv2.resize(cutout, (224, 224))  # BGR
+                # cv2.imwrite('test%i.jpg' % j, cutout)
+
+                # BGR to RGB, to 3x416x416
+                im = im[:, :, ::-1].transpose(2, 0, 1)
+                im = np.ascontiguousarray(
+                    im, dtype=np.float32)  # uint8 to float32
+                im /= 255.0  # 0 - 255 to 0.0 - 1.0
+                ims.append(im)
+
+            pred_cls2 = model(torch.Tensor(ims).to(d.device)
+                              ).argmax(1)  # classifier prediction
+            # retain matching class detections
+            x[i] = x[i][pred_cls1 == pred_cls2]
+
+    return x
+
+
+def increment_path(path, exist_ok=True, sep=''):
+    # Increment path, i.e. runs/exp --> runs/exp{sep}0, runs/exp{sep}1 etc.
+    path = Path(path)  # os-agnostic
+    if (path.exists() and exist_ok) or (not path.exists()):
+        return str(path)
+    else:
+        dirs = glob.glob(f"{path}{sep}*")  # similar paths
+        matches = [re.search(rf"%s{sep}(\d+)" % path.stem, d) for d in dirs]
+        i = [int(m.groups()[0]) for m in matches if m]  # indices
+        n = max(i) + 1 if i else 2  # increment number
+        return f"{path}{sep}{n}"  # update path

utils/google_app_engine/additional_requirements.txt

@@ -0,0 +1,4 @@
+# add these requirements in your app on top of the existing ones
+pip==18.1
+Flask==1.0.2
+gunicorn==19.9.0

utils/google_app_engine/app.yaml

@@ -0,0 +1,14 @@
+runtime: custom
+env: flex
+
+service: yolov5app
+
+liveness_check:
+  initial_delay_sec: 600
+
+manual_scaling:
+  instances: 1
+resources:
+  cpu: 1
+  memory_gb: 4
+  disk_size_gb: 20

utils/google_utils.py

@@ -0,0 +1,122 @@
+# Google utils: https://cloud.google.com/storage/docs/reference/libraries
+
+import os
+import platform
+import subprocess
+import time
+from pathlib import Path
+
+import requests
+import torch
+
+
+def gsutil_getsize(url=''):
+    # gs://bucket/file size https://cloud.google.com/storage/docs/gsutil/commands/du
+    s = subprocess.check_output(f'gsutil du {url}', shell=True).decode('utf-8')
+    return eval(s.split(' ')[0]) if len(s) else 0  # bytes
+
+
+def attempt_download(file, repo='ultralytics/yolov5'):
+    # Attempt file download if does not exist
+    file = Path(str(file).strip().replace("'", '').lower())
+
+    if not file.exists():
+        try:
+            response = requests.get(f'https://api.github.com/repos/{repo}/releases/latest').json()  # github api
+            assets = [x['name'] for x in response['assets']]  # release assets, i.e. ['yolov5s.pt', 'yolov5m.pt', ...]
+            tag = response['tag_name']  # i.e. 'v1.0'
+        except:  # fallback plan
+            assets = ['yolov5.pt', 'yolov5.pt', 'yolov5l.pt', 'yolov5x.pt']
+            tag = subprocess.check_output('git tag', shell=True).decode('utf-8').split('\n')[-2]
+
+        name = file.name
+        if name in assets:
+            msg = f'{file} missing, try downloading from https://github.com/{repo}/releases/'
+            redundant = False  # second download option
+            try:  # GitHub
+                url = f'https://github.com/{repo}/releases/download/{tag}/{name}'
+                print(f'Downloading {url} to {file}...')
+                torch.hub.download_url_to_file(url, file)
+                assert file.exists() and file.stat().st_size > 1E6  # check
+            except Exception as e:  # GCP
+                print(f'Download error: {e}')
+                assert redundant, 'No secondary mirror'
+                url = f'https://storage.googleapis.com/{repo}/ckpt/{name}'
+                print(f'Downloading {url} to {file}...')
+                os.system(f'curl -L {url} -o {file}')  # torch.hub.download_url_to_file(url, weights)
+            finally:
+                if not file.exists() or file.stat().st_size < 1E6:  # check
+                    file.unlink(missing_ok=True)  # remove partial downloads
+                    print(f'ERROR: Download failure: {msg}')
+                print('')
+                return
+
+
+def gdrive_download(id='16TiPfZj7htmTyhntwcZyEEAejOUxuT6m', file='tmp.zip'):
+    # Downloads a file from Google Drive. from yolov5.utils.google_utils import *; gdrive_download()
+    t = time.time()
+    file = Path(file)
+    cookie = Path('cookie')  # gdrive cookie
+    print(f'Downloading https://drive.google.com/uc?export=download&id={id} as {file}... ', end='')
+    file.unlink(missing_ok=True)  # remove existing file
+    cookie.unlink(missing_ok=True)  # remove existing cookie
+
+    # Attempt file download
+    out = "NUL" if platform.system() == "Windows" else "/dev/null"
+    os.system(f'curl -c ./cookie -s -L "drive.google.com/uc?export=download&id={id}" > {out}')
+    if os.path.exists('cookie'):  # large file
+        s = f'curl -Lb ./cookie "drive.google.com/uc?export=download&confirm={get_token()}&id={id}" -o {file}'
+    else:  # small file
+        s = f'curl -s -L -o {file} "drive.google.com/uc?export=download&id={id}"'
+    r = os.system(s)  # execute, capture return
+    cookie.unlink(missing_ok=True)  # remove existing cookie
+
+    # Error check
+    if r != 0:
+        file.unlink(missing_ok=True)  # remove partial
+        print('Download error ')  # raise Exception('Download error')
+        return r
+
+    # Unzip if archive
+    if file.suffix == '.zip':
+        print('unzipping... ', end='')
+        os.system(f'unzip -q {file}')  # unzip
+        file.unlink()  # remove zip to free space
+
+    print(f'Done ({time.time() - t:.1f}s)')
+    return r
+
+
+def get_token(cookie="./cookie"):
+    with open(cookie) as f:
+        for line in f:
+            if "download" in line:
+                return line.split()[-1]
+    return ""
+
+# def upload_blob(bucket_name, source_file_name, destination_blob_name):
+#     # Uploads a file to a bucket
+#     # https://cloud.google.com/storage/docs/uploading-objects#storage-upload-object-python
+#
+#     storage_client = storage.Client()
+#     bucket = storage_client.get_bucket(bucket_name)
+#     blob = bucket.blob(destination_blob_name)
+#
+#     blob.upload_from_filename(source_file_name)
+#
+#     print('File {} uploaded to {}.'.format(
+#         source_file_name,
+#         destination_blob_name))
+#
+#
+# def download_blob(bucket_name, source_blob_name, destination_file_name):
+#     # Uploads a blob from a bucket
+#     storage_client = storage.Client()
+#     bucket = storage_client.get_bucket(bucket_name)
+#     blob = bucket.blob(source_blob_name)
+#
+#     blob.download_to_filename(destination_file_name)
+#
+#     print('Blob {} downloaded to {}.'.format(
+#         source_blob_name,
+#         destination_file_name))

utils/infer_utils.py

@@ -0,0 +1,36 @@
+import torch
+
+
+
+def decode_infer(output, stride):
+    # logging.info(torch.tensor(output.shape[0]))
+    # logging.info(output.shape)
+    # # bz is batch-size
+    # bz = tuple(torch.tensor(output.shape[0]))
+    # gridsize = tuple(torch.tensor(output.shape[-1]))
+    # logging.info(gridsize)
+    sh = torch.tensor(output.shape)
+    bz = sh[0]
+    gridsize = sh[-1]
+
+    output = output.permute(0, 2, 3, 1)
+    output = output.view(bz, gridsize, gridsize, self.gt_per_grid, 5+self.numclass)
+    x1y1, x2y2, conf, prob = torch.split(
+        output, [2, 2, 1, self.numclass], dim=4)
+
+    shiftx = torch.arange(0, gridsize, dtype=torch.float32)
+    shifty = torch.arange(0, gridsize, dtype=torch.float32)
+    shifty, shiftx = torch.meshgrid([shiftx, shifty])
+    shiftx = shiftx.unsqueeze(-1).repeat(bz, 1, 1, self.gt_per_grid)
+    shifty = shifty.unsqueeze(-1).repeat(bz, 1, 1, self.gt_per_grid)
+
+    xy_grid = torch.stack([shiftx, shifty], dim=4).cuda()
+    x1y1 = (xy_grid+0.5-torch.exp(x1y1))*stride
+    x2y2 = (xy_grid+0.5+torch.exp(x2y2))*stride
+
+    xyxy = torch.cat((x1y1, x2y2), dim=4)
+    conf = torch.sigmoid(conf)
+    prob = torch.sigmoid(prob)
+    output = torch.cat((xyxy, conf, prob), 4)
+    output = output.view(bz, -1, 5+self.numclass)
+    return output

utils/loss.py

@@ -0,0 +1,304 @@
+# Loss functions
+
+import torch
+import torch.nn as nn
+import numpy as np
+from utils.general import bbox_iou
+from utils.torch_utils import is_parallel
+
+
+def smooth_BCE(eps=0.1):  # https://github.com/ultralytics/yolov3/issues/238#issuecomment-598028441
+    # return positive, negative label smoothing BCE targets
+    return 1.0 - 0.5 * eps, 0.5 * eps
+
+
+class BCEBlurWithLogitsLoss(nn.Module):
+    # BCEwithLogitLoss() with reduced missing label effects.
+    def __init__(self, alpha=0.05):
+        super(BCEBlurWithLogitsLoss, self).__init__()
+        self.loss_fcn = nn.BCEWithLogitsLoss(reduction='none')  # must be nn.BCEWithLogitsLoss()
+        self.alpha = alpha
+
+    def forward(self, pred, true):
+        loss = self.loss_fcn(pred, true)
+        pred = torch.sigmoid(pred)  # prob from logits
+        dx = pred - true  # reduce only missing label effects
+        # dx = (pred - true).abs()  # reduce missing label and false label effects
+        alpha_factor = 1 - torch.exp((dx - 1) / (self.alpha + 1e-4))
+        loss *= alpha_factor
+        return loss.mean()
+
+
+class FocalLoss(nn.Module):
+    # Wraps focal loss around existing loss_fcn(), i.e. criteria = FocalLoss(nn.BCEWithLogitsLoss(), gamma=1.5)
+    def __init__(self, loss_fcn, gamma=1.5, alpha=0.25):
+        super(FocalLoss, self).__init__()
+        self.loss_fcn = loss_fcn  # must be nn.BCEWithLogitsLoss()
+        self.gamma = gamma
+        self.alpha = alpha
+        self.reduction = loss_fcn.reduction
+        self.loss_fcn.reduction = 'none'  # required to apply FL to each element
+
+    def forward(self, pred, true):
+        loss = self.loss_fcn(pred, true)
+        # p_t = torch.exp(-loss)
+        # loss *= self.alpha * (1.000001 - p_t) ** self.gamma  # non-zero power for gradient stability
+
+        # TF implementation https://github.com/tensorflow/addons/blob/v0.7.1/tensorflow_addons/losses/focal_loss.py
+        pred_prob = torch.sigmoid(pred)  # prob from logits
+        p_t = true * pred_prob + (1 - true) * (1 - pred_prob)
+        alpha_factor = true * self.alpha + (1 - true) * (1 - self.alpha)
+        modulating_factor = (1.0 - p_t) ** self.gamma
+        loss *= alpha_factor * modulating_factor
+
+        if self.reduction == 'mean':
+            return loss.mean()
+        elif self.reduction == 'sum':
+            return loss.sum()
+        else:  # 'none'
+            return loss
+
+
+class QFocalLoss(nn.Module):
+    # Wraps Quality focal loss around existing loss_fcn(), i.e. criteria = FocalLoss(nn.BCEWithLogitsLoss(), gamma=1.5)
+    def __init__(self, loss_fcn, gamma=1.5, alpha=0.25):
+        super(QFocalLoss, self).__init__()
+        self.loss_fcn = loss_fcn  # must be nn.BCEWithLogitsLoss()
+        self.gamma = gamma
+        self.alpha = alpha
+        self.reduction = loss_fcn.reduction
+        self.loss_fcn.reduction = 'none'  # required to apply FL to each element
+
+    def forward(self, pred, true):
+        loss = self.loss_fcn(pred, true)
+
+        pred_prob = torch.sigmoid(pred)  # prob from logits
+        alpha_factor = true * self.alpha + (1 - true) * (1 - self.alpha)
+        modulating_factor = torch.abs(true - pred_prob) ** self.gamma
+        loss *= alpha_factor * modulating_factor
+
+        if self.reduction == 'mean':
+            return loss.mean()
+        elif self.reduction == 'sum':
+            return loss.sum()
+        else:  # 'none'
+            return loss
+
+class WingLoss(nn.Module):
+    def __init__(self, w=10, e=2):
+        super(WingLoss, self).__init__()
+        # https://arxiv.org/pdf/1711.06753v4.pdf   Figure 5
+        self.w = w
+        self.e = e
+        self.C = self.w - self.w * np.log(1 + self.w / self.e)
+
+    def forward(self, x, t, sigma=1):
+        weight = torch.ones_like(t)
+        weight[torch.where(t==-1)] = 0
+        diff = weight * (x - t)
+        abs_diff = diff.abs()
+        flag = (abs_diff.data < self.w).float()
+        y = flag * self.w * torch.log(1 + abs_diff / self.e) + (1 - flag) * (abs_diff - self.C)
+        return y.sum()
+
+class LandmarksLoss(nn.Module):
+    # BCEwithLogitLoss() with reduced missing label effects.
+    def __init__(self, alpha=1.0):
+        super(LandmarksLoss, self).__init__()
+        self.loss_fcn = WingLoss()#nn.SmoothL1Loss(reduction='sum')
+        self.alpha = alpha
+
+    def forward(self, pred, truel, mask):
+        loss = self.loss_fcn(pred*mask, truel*mask)
+        return loss / (torch.sum(mask) + 10e-14)
+
+
+def compute_loss(p, targets, model):  # predictions, targets, model
+    device = targets.device
+    lcls, lbox, lobj, lmark = torch.zeros(1, device=device), torch.zeros(1, device=device), torch.zeros(1, device=device), torch.zeros(1, device=device)
+    tcls, tbox, indices, anchors, tlandmarks, lmks_mask = build_targets(p, targets, model)  # targets
+    h = model.hyp  # hyperparameters
+
+    # Define criteria
+    BCEcls = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['cls_pw']], device=device))  # weight=model.class_weights)
+    BCEobj = nn.BCEWithLogitsLoss(pos_weight=torch.tensor([h['obj_pw']], device=device))
+
+    landmarks_loss = LandmarksLoss(1.0)
+
+    # Class label smoothing https://arxiv.org/pdf/1902.04103.pdf eqn 3
+    cp, cn = smooth_BCE(eps=0.0)
+
+    # Focal loss
+    g = h['fl_gamma']  # focal loss gamma
+    if g > 0:
+        BCEcls, BCEobj = FocalLoss(BCEcls, g), FocalLoss(BCEobj, g)
+
+    # Losses
+    nt = 0  # number of targets
+    no = len(p)  # number of outputs
+    balance = [4.0, 1.0, 0.4] if no == 3 else [4.0, 1.0, 0.4, 0.1]  # P3-5 or P3-6
+    for i, pi in enumerate(p):  # layer index, layer predictions
+        b, a, gj, gi = indices[i]  # image, anchor, gridy, gridx
+        tobj = torch.zeros_like(pi[..., 0], device=device)  # target obj
+
+        n = b.shape[0]  # number of targets
+        if n:
+            nt += n  # cumulative targets
+            ps = pi[b, a, gj, gi]  # prediction subset corresponding to targets
+
+            # Regression
+            pxy = ps[:, :2].sigmoid() * 2. - 0.5
+            pwh = (ps[:, 2:4].sigmoid() * 2) ** 2 * anchors[i]
+            pbox = torch.cat((pxy, pwh), 1)  # predicted box
+            iou = bbox_iou(pbox.T, tbox[i], x1y1x2y2=False, CIoU=True)  # iou(prediction, target)
+            lbox += (1.0 - iou).mean()  # iou loss
+
+            # Objectness
+            tobj[b, a, gj, gi] = (1.0 - model.gr) + model.gr * iou.detach().clamp(0).type(tobj.dtype)  # iou ratio
+
+            # Classification
+            if model.nc > 1:  # cls loss (only if multiple classes)
+                t = torch.full_like(ps[:, 13:], cn, device=device)  # targets
+                t[range(n), tcls[i]] = cp
+                lcls += BCEcls(ps[:, 13:], t)  # BCE
+
+            # Append targets to text file
+            # with open('targets.txt', 'a') as file:
+            #     [file.write('%11.5g ' * 4 % tuple(x) + '\n') for x in torch.cat((txy[i], twh[i]), 1)]
+
+            #landmarks loss
+            #plandmarks = ps[:,5:13].sigmoid() * 8. - 4.
+            plandmarks = ps[:,5:13]
+
+            plandmarks[:, 0:2] = plandmarks[:, 0:2] * anchors[i]
+            plandmarks[:, 2:4] = plandmarks[:, 2:4] * anchors[i]
+            plandmarks[:, 4:6] = plandmarks[:, 4:6] * anchors[i]
+            plandmarks[:, 6:8] = plandmarks[:, 6:8] * anchors[i]
+            # plandmarks[:, 8:10] = plandmarks[:,8:10] * anchors[i]
+
+            lmark += landmarks_loss(plandmarks, tlandmarks[i], lmks_mask[i])
+
+
+        lobj += BCEobj(pi[..., 4], tobj) * balance[i]  # obj loss
+
+    s = 3 / no  # output count scaling
+    lbox *= h['box'] * s
+    lobj *= h['obj'] * s * (1.4 if no == 4 else 1.)
+    lcls *= h['cls'] * s
+    lmark *= h['landmark'] * s
+
+    bs = tobj.shape[0]  # batch size
+
+    loss = lbox + lobj + lcls + lmark
+    return loss * bs, torch.cat((lbox, lobj, lcls, lmark, loss)).detach()
+
+
+def build_targets(p, targets, model):
+    # Build targets for compute_loss(), input targets(image,class,x,y,w,h)
+    det = model.module.model[-1] if is_parallel(model) else model.model[-1]  # Detect() module
+    na, nt = det.na, targets.shape[0]  # number of anchors, targets
+    tcls, tbox, indices, anch, landmarks, lmks_mask = [], [], [], [], [], []
+    #gain = torch.ones(7, device=targets.device)  # normalized to gridspace gain
+    gain = torch.ones(15, device=targets.device)
+    ai = torch.arange(na, device=targets.device).float().view(na, 1).repeat(1, nt)  # same as .repeat_interleave(nt)
+    targets = torch.cat((targets.repeat(na, 1, 1), ai[:, :, None]), 2)  # append anchor indices
+
+    g = 0.5  # bias
+    off = torch.tensor([[0, 0],
+                        [1, 0], [0, 1], [-1, 0], [0, -1],  # j,k,l,m
+                        # [1, 1], [1, -1], [-1, 1], [-1, -1],  # jk,jm,lk,lm
+                        ], device=targets.device).float() * g  # offsets
+
+    for i in range(det.nl):
+        anchors = det.anchors[i]
+        gain[2:6] = torch.tensor(p[i].shape)[[3, 2, 3, 2]]  # xyxy gain
+        #landmarks 10
+        gain[6:14] = torch.tensor(p[i].shape)[[3, 2, 3, 2, 3, 2, 3, 2]]  # xyxy gain
+
+        # Match targets to anchors
+        t = targets * gain
+        if nt:
+            # Matches
+            r = t[:, :, 4:6] / anchors[:, None]  # wh ratio
+            j = torch.max(r, 1. / r).max(2)[0] < model.hyp['anchor_t']  # compare
+            # j = wh_iou(anchors, t[:, 4:6]) > model.hyp['iou_t']  # iou(3,n)=wh_iou(anchors(3,2), gwh(n,2))
+            t = t[j]  # filter
+
+            # Offsets
+            gxy = t[:, 2:4]  # grid xy
+            gxi = gain[[2, 3]] - gxy  # inverse
+            j, k = ((gxy % 1. < g) & (gxy > 1.)).T
+            l, m = ((gxi % 1. < g) & (gxi > 1.)).T
+            j = torch.stack((torch.ones_like(j), j, k, l, m))
+            t = t.repeat((5, 1, 1))[j]
+            offsets = (torch.zeros_like(gxy)[None] + off[:, None])[j]
+        else:
+            t = targets[0]
+            offsets = 0
+
+        # Define
+        b, c = t[:, :2].long().T  # image, class
+        gxy = t[:, 2:4]  # grid xy
+        gwh = t[:, 4:6]  # grid wh
+        gij = (gxy - offsets).long()
+        gi, gj = gij.T  # grid xy indices
+
+        # Append
+        a = t[:, 14].long()  # anchor indices
+        indices.append((b, a, gj.clamp_(0, gain[3] - 1), gi.clamp_(0, gain[2] - 1)))  # image, anchor, grid indices
+        tbox.append(torch.cat((gxy - gij, gwh), 1))  # box
+        anch.append(anchors[a])  # anchors
+        tcls.append(c)  # class
+
+        #landmarks
+        lks = t[:,6:14]
+        #lks_mask = lks > 0
+        #lks_mask = lks_mask.float()
+        lks_mask = torch.where(lks < 0, torch.full_like(lks, 0.), torch.full_like(lks, 1.0))
+
+        #应该是关键点的坐标除以anch的宽高才对，便于模型学习。使用gwh会导致不同关键点的编码不同，没有统一的参考标准
+
+        lks[:, [0, 1]] = (lks[:, [0, 1]] - gij)
+        lks[:, [2, 3]] = (lks[:, [2, 3]] - gij)
+        lks[:, [4, 5]] = (lks[:, [4, 5]] - gij)
+        lks[:, [6, 7]] = (lks[:, [6, 7]] - gij)
+        # lks[:, [8, 9]] = (lks[:, [8, 9]] - gij)
+
+        '''
+        #anch_w = torch.ones(5, device=targets.device).fill_(anchors[0][0])
+        #anch_wh = torch.ones(5, device=targets.device)
+        anch_f_0 = (a == 0).unsqueeze(1).repeat(1, 5)
+        anch_f_1 = (a == 1).unsqueeze(1).repeat(1, 5)
+        anch_f_2 = (a == 2).unsqueeze(1).repeat(1, 5)
+        lks[:, [0, 2, 4, 6, 8]] = torch.where(anch_f_0, lks[:, [0, 2, 4, 6, 8]] / anchors[0][0], lks[:, [0, 2, 4, 6, 8]])
+        lks[:, [0, 2, 4, 6, 8]] = torch.where(anch_f_1, lks[:, [0, 2, 4, 6, 8]] / anchors[1][0], lks[:, [0, 2, 4, 6, 8]])
+        lks[:, [0, 2, 4, 6, 8]] = torch.where(anch_f_2, lks[:, [0, 2, 4, 6, 8]] / anchors[2][0], lks[:, [0, 2, 4, 6, 8]])
+
+        lks[:, [1, 3, 5, 7, 9]] = torch.where(anch_f_0, lks[:, [1, 3, 5, 7, 9]] / anchors[0][1], lks[:, [1, 3, 5, 7, 9]])
+        lks[:, [1, 3, 5, 7, 9]] = torch.where(anch_f_1, lks[:, [1, 3, 5, 7, 9]] / anchors[1][1], lks[:, [1, 3, 5, 7, 9]])
+        lks[:, [1, 3, 5, 7, 9]] = torch.where(anch_f_2, lks[:, [1, 3, 5, 7, 9]] / anchors[2][1], lks[:, [1, 3, 5, 7, 9]])
+
+        #new_lks = lks[lks_mask>0]
+        #print('new_lks:   min --- ', torch.min(new_lks), '  max --- ', torch.max(new_lks))
+        
+        lks_mask_1 = torch.where(lks < -3, torch.full_like(lks, 0.), torch.full_like(lks, 1.0))
+        lks_mask_2 = torch.where(lks > 3, torch.full_like(lks, 0.), torch.full_like(lks, 1.0))
+
+        lks_mask_new = lks_mask * lks_mask_1 * lks_mask_2
+        lks_mask_new[:, 0] = lks_mask_new[:, 0] * lks_mask_new[:, 1]
+        lks_mask_new[:, 1] = lks_mask_new[:, 0] * lks_mask_new[:, 1]
+        lks_mask_new[:, 2] = lks_mask_new[:, 2] * lks_mask_new[:, 3]
+        lks_mask_new[:, 3] = lks_mask_new[:, 2] * lks_mask_new[:, 3]
+        lks_mask_new[:, 4] = lks_mask_new[:, 4] * lks_mask_new[:, 5]
+        lks_mask_new[:, 5] = lks_mask_new[:, 4] * lks_mask_new[:, 5]
+        lks_mask_new[:, 6] = lks_mask_new[:, 6] * lks_mask_new[:, 7]
+        lks_mask_new[:, 7] = lks_mask_new[:, 6] * lks_mask_new[:, 7]
+        lks_mask_new[:, 8] = lks_mask_new[:, 8] * lks_mask_new[:, 9]
+        lks_mask_new[:, 9] = lks_mask_new[:, 8] * lks_mask_new[:, 9]
+        '''
+        lks_mask_new = lks_mask
+        lmks_mask.append(lks_mask_new)
+        landmarks.append(lks)
+        #print('lks: ',  lks.size())
+
+    return tcls, tbox, indices, anch, landmarks, lmks_mask

utils/metrics.py

@@ -0,0 +1,200 @@
+# Model validation metrics
+
+from pathlib import Path
+
+import matplotlib.pyplot as plt
+import numpy as np
+import torch
+
+from . import general
+
+
+def fitness(x):
+    # Model fitness as a weighted combination of metrics
+    w = [0.0, 0.0, 0.1, 0.9]  # weights for [P, R, mAP@0.5, mAP@0.5:0.95]
+    return (x[:, :4] * w).sum(1)
+
+
+def ap_per_class(tp, conf, pred_cls, target_cls, plot=False, save_dir='precision-recall_curve.png', names=[]):
+    """ Compute the average precision, given the recall and precision curves.
+    Source: https://github.com/rafaelpadilla/Object-Detection-Metrics.
+    # Arguments
+        tp:  True positives (nparray, nx1 or nx10).
+        conf:  Objectness value from 0-1 (nparray).
+        pred_cls:  Predicted object classes (nparray).
+        target_cls:  True object classes (nparray).
+        plot:  Plot precision-recall curve at mAP@0.5
+        save_dir:  Plot save directory
+    # Returns
+        The average precision as computed in py-faster-rcnn.
+    """
+
+    # Sort by objectness
+    i = np.argsort(-conf)
+    tp, conf, pred_cls = tp[i], conf[i], pred_cls[i]
+
+    # Find unique classes
+    unique_classes = np.unique(target_cls)
+
+    # Create Precision-Recall curve and compute AP for each class
+    px, py = np.linspace(0, 1, 1000), []  # for plotting
+    pr_score = 0.1  # score to evaluate P and R https://github.com/ultralytics/yolov3/issues/898
+    s = [unique_classes.shape[0], tp.shape[1]]  # number class, number iou thresholds (i.e. 10 for mAP0.5...0.95)
+    ap, p, r = np.zeros(s), np.zeros(s), np.zeros(s)
+    for ci, c in enumerate(unique_classes):
+        i = pred_cls == c
+        n_l = (target_cls == c).sum()  # number of labels
+        n_p = i.sum()  # number of predictions
+
+        if n_p == 0 or n_l == 0:
+            continue
+        else:
+            # Accumulate FPs and TPs
+            fpc = (1 - tp[i]).cumsum(0)
+            tpc = tp[i].cumsum(0)
+
+            # Recall
+            recall = tpc / (n_l + 1e-16)  # recall curve
+            r[ci] = np.interp(-pr_score, -conf[i], recall[:, 0])  # r at pr_score, negative x, xp because xp decreases
+
+            # Precision
+            precision = tpc / (tpc + fpc)  # precision curve
+            p[ci] = np.interp(-pr_score, -conf[i], precision[:, 0])  # p at pr_score
+
+            # AP from recall-precision curve
+            for j in range(tp.shape[1]):
+                ap[ci, j], mpre, mrec = compute_ap(recall[:, j], precision[:, j])
+                if plot and (j == 0):
+                    py.append(np.interp(px, mrec, mpre))  # precision at mAP@0.5
+
+    # Compute F1 score (harmonic mean of precision and recall)
+    f1 = 2 * p * r / (p + r + 1e-16)
+
+    if plot:
+        plot_pr_curve(px, py, ap, save_dir, names)
+
+    return p, r, ap, f1, unique_classes.astype('int32')
+
+
+def compute_ap(recall, precision):
+    """ Compute the average precision, given the recall and precision curves
+    # Arguments
+        recall:    The recall curve (list)
+        precision: The precision curve (list)
+    # Returns
+        Average precision, precision curve, recall curve
+    """
+
+    # Append sentinel values to beginning and end
+    mrec = np.concatenate(([0.], recall, [recall[-1] + 0.01]))
+    mpre = np.concatenate(([1.], precision, [0.]))
+
+    # Compute the precision envelope
+    mpre = np.flip(np.maximum.accumulate(np.flip(mpre)))
+
+    # Integrate area under curve
+    method = 'interp'  # methods: 'continuous', 'interp'
+    if method == 'interp':
+        x = np.linspace(0, 1, 101)  # 101-point interp (COCO)
+        ap = np.trapz(np.interp(x, mrec, mpre), x)  # integrate
+    else:  # 'continuous'
+        i = np.where(mrec[1:] != mrec[:-1])[0]  # points where x axis (recall) changes
+        ap = np.sum((mrec[i + 1] - mrec[i]) * mpre[i + 1])  # area under curve
+
+    return ap, mpre, mrec
+
+
+class ConfusionMatrix:
+    # Updated version of https://github.com/kaanakan/object_detection_confusion_matrix
+    def __init__(self, nc, conf=0.25, iou_thres=0.45):
+        self.matrix = np.zeros((nc + 1, nc + 1))
+        self.nc = nc  # number of classes
+        self.conf = conf
+        self.iou_thres = iou_thres
+
+    def process_batch(self, detections, labels):
+        """
+        Return intersection-over-union (Jaccard index) of boxes.
+        Both sets of boxes are expected to be in (x1, y1, x2, y2) format.
+        Arguments:
+            detections (Array[N, 6]), x1, y1, x2, y2, conf, class
+            labels (Array[M, 5]), class, x1, y1, x2, y2
+        Returns:
+            None, updates confusion matrix accordingly
+        """
+        detections = detections[detections[:, 4] > self.conf]
+        gt_classes = labels[:, 0].int()
+        detection_classes = detections[:, 5].int()
+        iou = general.box_iou(labels[:, 1:], detections[:, :4])
+
+        x = torch.where(iou > self.iou_thres)
+        if x[0].shape[0]:
+            matches = torch.cat((torch.stack(x, 1), iou[x[0], x[1]][:, None]), 1).cpu().numpy()
+            if x[0].shape[0] > 1:
+                matches = matches[matches[:, 2].argsort()[::-1]]
+                matches = matches[np.unique(matches[:, 1], return_index=True)[1]]
+                matches = matches[matches[:, 2].argsort()[::-1]]
+                matches = matches[np.unique(matches[:, 0], return_index=True)[1]]
+        else:
+            matches = np.zeros((0, 3))
+
+        n = matches.shape[0] > 0
+        m0, m1, _ = matches.transpose().astype(np.int16)
+        for i, gc in enumerate(gt_classes):
+            j = m0 == i
+            if n and sum(j) == 1:
+                self.matrix[gc, detection_classes[m1[j]]] += 1  # correct
+            else:
+                self.matrix[gc, self.nc] += 1  # background FP
+
+        if n:
+            for i, dc in enumerate(detection_classes):
+                if not any(m1 == i):
+                    self.matrix[self.nc, dc] += 1  # background FN
+
+    def matrix(self):
+        return self.matrix
+
+    def plot(self, save_dir='', names=()):
+        try:
+            import seaborn as sn
+
+            array = self.matrix / (self.matrix.sum(0).reshape(1, self.nc + 1) + 1E-6)  # normalize
+            array[array < 0.005] = np.nan  # don't annotate (would appear as 0.00)
+
+            fig = plt.figure(figsize=(12, 9), tight_layout=True)
+            sn.set(font_scale=1.0 if self.nc < 50 else 0.8)  # for label size
+            labels = (0 < len(names) < 99) and len(names) == self.nc  # apply names to ticklabels
+            sn.heatmap(array, annot=self.nc < 30, annot_kws={"size": 8}, cmap='Blues', fmt='.2f', square=True,
+                       xticklabels=names + ['background FN'] if labels else "auto",
+                       yticklabels=names + ['background FP'] if labels else "auto").set_facecolor((1, 1, 1))
+            fig.axes[0].set_xlabel('True')
+            fig.axes[0].set_ylabel('Predicted')
+            fig.savefig(Path(save_dir) / 'confusion_matrix.png', dpi=250)
+        except Exception as e:
+            pass
+
+    def print(self):
+        for i in range(self.nc + 1):
+            print(' '.join(map(str, self.matrix[i])))
+
+
+# Plots ----------------------------------------------------------------------------------------------------------------
+
+def plot_pr_curve(px, py, ap, save_dir='.', names=()):
+    fig, ax = plt.subplots(1, 1, figsize=(9, 6), tight_layout=True)
+    py = np.stack(py, axis=1)
+
+    if 0 < len(names) < 21:  # show mAP in legend if < 10 classes
+        for i, y in enumerate(py.T):
+            ax.plot(px, y, linewidth=1, label=f'{names[i]} %.3f' % ap[i, 0])  # plot(recall, precision)
+    else:
+        ax.plot(px, py, linewidth=1, color='grey')  # plot(recall, precision)
+
+    ax.plot(px, py.mean(1), linewidth=3, color='blue', label='all classes %.3f mAP@0.5' % ap[:, 0].mean())
+    ax.set_xlabel('Recall')
+    ax.set_ylabel('Precision')
+    ax.set_xlim(0, 1)
+    ax.set_ylim(0, 1)
+    plt.legend(bbox_to_anchor=(1.04, 1), loc="upper left")
+    fig.savefig(Path(save_dir) / 'precision_recall_curve.png', dpi=250)

utils/plots.py

@@ -0,0 +1,413 @@
+# Plotting utils
+
+import glob
+import math
+import os
+import random
+from copy import copy
+from pathlib import Path
+
+import cv2
+import matplotlib
+import matplotlib.pyplot as plt
+import numpy as np
+import pandas as pd
+import seaborn as sns
+import torch
+import yaml
+from PIL import Image, ImageDraw
+from scipy.signal import butter, filtfilt
+
+from utils.general import xywh2xyxy, xyxy2xywh
+from utils.metrics import fitness
+
+# Settings
+matplotlib.rc('font', **{'size': 11})
+matplotlib.use('Agg')  # for writing to files only
+
+
+def color_list():
+    # Return first 10 plt colors as (r,g,b) https://stackoverflow.com/questions/51350872/python-from-color-name-to-rgb
+    def hex2rgb(h):
+        return tuple(int(h[1 + i:1 + i + 2], 16) for i in (0, 2, 4))
+
+    return [hex2rgb(h) for h in plt.rcParams['axes.prop_cycle'].by_key()['color']]
+
+
+def hist2d(x, y, n=100):
+    # 2d histogram used in labels.png and evolve.png
+    xedges, yedges = np.linspace(x.min(), x.max(), n), np.linspace(y.min(), y.max(), n)
+    hist, xedges, yedges = np.histogram2d(x, y, (xedges, yedges))
+    xidx = np.clip(np.digitize(x, xedges) - 1, 0, hist.shape[0] - 1)
+    yidx = np.clip(np.digitize(y, yedges) - 1, 0, hist.shape[1] - 1)
+    return np.log(hist[xidx, yidx])
+
+
+def butter_lowpass_filtfilt(data, cutoff=1500, fs=50000, order=5):
+    # https://stackoverflow.com/questions/28536191/how-to-filter-smooth-with-scipy-numpy
+    def butter_lowpass(cutoff, fs, order):
+        nyq = 0.5 * fs
+        normal_cutoff = cutoff / nyq
+        return butter(order, normal_cutoff, btype='low', analog=False)
+
+    b, a = butter_lowpass(cutoff, fs, order=order)
+    return filtfilt(b, a, data)  # forward-backward filter
+
+
+def plot_one_box(x, img, color=None, label=None, line_thickness=None):
+    # Plots one bounding box on image img
+    tl = line_thickness or round(0.002 * (img.shape[0] + img.shape[1]) / 2) + 1  # line/font thickness
+    color = color or [random.randint(0, 255) for _ in range(3)]
+    c1, c2 = (int(x[0]), int(x[1])), (int(x[2]), int(x[3]))
+    cv2.rectangle(img, c1, c2, color, thickness=tl, lineType=cv2.LINE_AA)
+    if label:
+        tf = max(tl - 1, 1)  # font thickness
+        t_size = cv2.getTextSize(label, 0, fontScale=tl / 3, thickness=tf)[0]
+        c2 = c1[0] + t_size[0], c1[1] - t_size[1] - 3
+        cv2.rectangle(img, c1, c2, color, -1, cv2.LINE_AA)  # filled
+        cv2.putText(img, label, (c1[0], c1[1] - 2), 0, tl / 3, [225, 255, 255], thickness=tf, lineType=cv2.LINE_AA)
+
+
+def plot_wh_methods():  # from utils.plots import *; plot_wh_methods()
+    # Compares the two methods for width-height anchor multiplication
+    # https://github.com/ultralytics/yolov3/issues/168
+    x = np.arange(-4.0, 4.0, .1)
+    ya = np.exp(x)
+    yb = torch.sigmoid(torch.from_numpy(x)).numpy() * 2
+
+    fig = plt.figure(figsize=(6, 3), tight_layout=True)
+    plt.plot(x, ya, '.-', label='YOLOv3')
+    plt.plot(x, yb ** 2, '.-', label='YOLOv5 ^2')
+    plt.plot(x, yb ** 1.6, '.-', label='YOLOv5 ^1.6')
+    plt.xlim(left=-4, right=4)
+    plt.ylim(bottom=0, top=6)
+    plt.xlabel('input')
+    plt.ylabel('output')
+    plt.grid()
+    plt.legend()
+    fig.savefig('comparison.png', dpi=200)
+
+
+def output_to_target(output):
+    # Convert model output to target format [batch_id, class_id, x, y, w, h, conf]
+    targets = []
+    for i, o in enumerate(output):
+        for *box, conf, cls in o.cpu().numpy():
+            targets.append([i, cls, *list(*xyxy2xywh(np.array(box)[None])), conf])
+    return np.array(targets)
+
+
+def plot_images(images, targets, paths=None, fname='images.jpg', names=None, max_size=640, max_subplots=16):
+    # Plot image grid with labels
+
+    if isinstance(images, torch.Tensor):
+        images = images.cpu().float().numpy()
+    if isinstance(targets, torch.Tensor):
+        targets = targets.cpu().numpy()
+
+    # un-normalise
+    if np.max(images[0]) <= 1:
+        images *= 255
+
+    tl = 3  # line thickness
+    tf = max(tl - 1, 1)  # font thickness
+    bs, _, h, w = images.shape  # batch size, _, height, width
+    bs = min(bs, max_subplots)  # limit plot images
+    ns = np.ceil(bs ** 0.5)  # number of subplots (square)
+
+    # Check if we should resize
+    scale_factor = max_size / max(h, w)
+    if scale_factor < 1:
+        h = math.ceil(scale_factor * h)
+        w = math.ceil(scale_factor * w)
+
+    # colors = color_list()  # list of colors
+    mosaic = np.full((int(ns * h), int(ns * w), 3), 255, dtype=np.uint8)  # init
+    for i, img in enumerate(images):
+        if i == max_subplots:  # if last batch has fewer images than we expect
+            break
+
+        block_x = int(w * (i // ns))
+        block_y = int(h * (i % ns))
+
+        img = img.transpose(1, 2, 0)
+        if scale_factor < 1:
+            img = cv2.resize(img, (w, h))
+
+        mosaic[block_y:block_y + h, block_x:block_x + w, :] = img
+        if len(targets) > 0:
+            image_targets = targets[targets[:, 0] == i]
+            boxes = xywh2xyxy(image_targets[:, 2:6]).T
+            classes = image_targets[:, 1].astype('int')
+            labels = image_targets.shape[1] == 6  # labels if no conf column
+            conf = None if labels else image_targets[:, 6]  # check for confidence presence (label vs pred)
+
+            if boxes.shape[1]:
+                if boxes.max() <= 1.01:  # if normalized with tolerance 0.01
+                    boxes[[0, 2]] *= w  # scale to pixels
+                    boxes[[1, 3]] *= h
+                elif scale_factor < 1:  # absolute coords need scale if image scales
+                    boxes *= scale_factor
+            boxes[[0, 2]] += block_x
+            boxes[[1, 3]] += block_y
+            for j, box in enumerate(boxes.T):
+                cls = int(classes[j])
+                # color = colors[cls % len(colors)]
+                cls = names[cls] if names else cls
+                if labels or conf[j] > 0.25:  # 0.25 conf thresh
+                    label = '%s' % cls if labels else '%s %.1f' % (cls, conf[j])
+                    plot_one_box(box, mosaic, label=label, color=None, line_thickness=tl)
+
+        # Draw image filename labels
+        if paths:
+            label = Path(paths[i]).name[:40]  # trim to 40 char
+            t_size = cv2.getTextSize(label, 0, fontScale=tl / 3, thickness=tf)[0]
+            cv2.putText(mosaic, label, (block_x + 5, block_y + t_size[1] + 5), 0, tl / 3, [220, 220, 220], thickness=tf,
+                        lineType=cv2.LINE_AA)
+
+        # Image border
+        cv2.rectangle(mosaic, (block_x, block_y), (block_x + w, block_y + h), (255, 255, 255), thickness=3)
+
+    if fname:
+        r = min(1280. / max(h, w) / ns, 1.0)  # ratio to limit image size
+        mosaic = cv2.resize(mosaic, (int(ns * w * r), int(ns * h * r)), interpolation=cv2.INTER_AREA)
+        # cv2.imwrite(fname, cv2.cvtColor(mosaic, cv2.COLOR_BGR2RGB))  # cv2 save
+        Image.fromarray(mosaic).save(fname)  # PIL save
+    return mosaic
+
+
+def plot_lr_scheduler(optimizer, scheduler, epochs=300, save_dir=''):
+    # Plot LR simulating training for full epochs
+    optimizer, scheduler = copy(optimizer), copy(scheduler)  # do not modify originals
+    y = []
+    for _ in range(epochs):
+        scheduler.step()
+        y.append(optimizer.param_groups[0]['lr'])
+    plt.plot(y, '.-', label='LR')
+    plt.xlabel('epoch')
+    plt.ylabel('LR')
+    plt.grid()
+    plt.xlim(0, epochs)
+    plt.ylim(0)
+    plt.savefig(Path(save_dir) / 'LR.png', dpi=200)
+    plt.close()
+
+
+def plot_test_txt():  # from utils.plots import *; plot_test()
+    # Plot test.txt histograms
+    x = np.loadtxt('test.txt', dtype=np.float32)
+    box = xyxy2xywh(x[:, :4])
+    cx, cy = box[:, 0], box[:, 1]
+
+    fig, ax = plt.subplots(1, 1, figsize=(6, 6), tight_layout=True)
+    ax.hist2d(cx, cy, bins=600, cmax=10, cmin=0)
+    ax.set_aspect('equal')
+    plt.savefig('hist2d.png', dpi=300)
+
+    fig, ax = plt.subplots(1, 2, figsize=(12, 6), tight_layout=True)
+    ax[0].hist(cx, bins=600)
+    ax[1].hist(cy, bins=600)
+    plt.savefig('hist1d.png', dpi=200)
+
+
+def plot_targets_txt():  # from utils.plots import *; plot_targets_txt()
+    # Plot targets.txt histograms
+    x = np.loadtxt('targets.txt', dtype=np.float32).T
+    s = ['x targets', 'y targets', 'width targets', 'height targets']
+    fig, ax = plt.subplots(2, 2, figsize=(8, 8), tight_layout=True)
+    ax = ax.ravel()
+    for i in range(4):
+        ax[i].hist(x[i], bins=100, label='%.3g +/- %.3g' % (x[i].mean(), x[i].std()))
+        ax[i].legend()
+        ax[i].set_title(s[i])
+    plt.savefig('targets.jpg', dpi=200)
+
+
+def plot_study_txt(path='study/', x=None):  # from utils.plots import *; plot_study_txt()
+    # Plot study.txt generated by test.py
+    fig, ax = plt.subplots(2, 4, figsize=(10, 6), tight_layout=True)
+    ax = ax.ravel()
+
+    fig2, ax2 = plt.subplots(1, 1, figsize=(8, 4), tight_layout=True)
+    for f in [Path(path) / f'study_coco_{x}.txt' for x in ['yolov5s', 'yolov5m', 'yolov5l', 'yolov5x']]:
+        y = np.loadtxt(f, dtype=np.float32, usecols=[0, 1, 2, 3, 7, 8, 9], ndmin=2).T
+        x = np.arange(y.shape[1]) if x is None else np.array(x)
+        s = ['P', 'R', 'mAP@.5', 'mAP@.5:.95', 't_inference (ms/img)', 't_NMS (ms/img)', 't_total (ms/img)']
+        for i in range(7):
+            ax[i].plot(x, y[i], '.-', linewidth=2, markersize=8)
+            ax[i].set_title(s[i])
+
+        j = y[3].argmax() + 1
+        ax2.plot(y[6, :j], y[3, :j] * 1E2, '.-', linewidth=2, markersize=8,
+                 label=f.stem.replace('study_coco_', '').replace('yolo', 'YOLO'))
+
+    ax2.plot(1E3 / np.array([209, 140, 97, 58, 35, 18]), [34.6, 40.5, 43.0, 47.5, 49.7, 51.5],
+             'k.-', linewidth=2, markersize=8, alpha=.25, label='EfficientDet')
+
+    ax2.grid()
+    ax2.set_yticks(np.arange(30, 60, 5))
+    ax2.set_xlim(0, 30)
+    ax2.set_ylim(29, 51)
+    ax2.set_xlabel('GPU Speed (ms/img)')
+    ax2.set_ylabel('COCO AP val')
+    ax2.legend(loc='lower right')
+    plt.savefig('test_study.png', dpi=300)
+
+
+def plot_labels(labels, save_dir=Path(''), loggers=None):
+    # plot dataset labels
+    print('Plotting labels... ')
+    c, b = labels[:, 0], labels[:, 1:5].transpose()  # classes, boxes
+    nc = int(c.max() + 1)  # number of classes
+    colors = color_list()
+    x = pd.DataFrame(b.transpose(), columns=['x', 'y', 'width', 'height'])
+
+    # seaborn correlogram
+    sns.pairplot(x, corner=True, diag_kind='auto', kind='hist', diag_kws=dict(bins=50), plot_kws=dict(pmax=0.9))
+    plt.savefig(save_dir / 'labels_correlogram.jpg', dpi=200)
+    plt.close()
+
+    # matplotlib labels
+    matplotlib.use('svg')  # faster
+    ax = plt.subplots(2, 2, figsize=(8, 8), tight_layout=True)[1].ravel()
+    ax[0].hist(c, bins=np.linspace(0, nc, nc + 1) - 0.5, rwidth=0.8)
+    ax[0].set_xlabel('classes')
+    sns.histplot(x, x='x', y='y', ax=ax[2], bins=50, pmax=0.9)
+    sns.histplot(x, x='width', y='height', ax=ax[3], bins=50, pmax=0.9)
+
+    # rectangles
+    labels[:, 1:3] = 0.5  # center
+    labels[:, 1:] = xywh2xyxy(labels[:, 1:]) * 2000
+    img = Image.fromarray(np.ones((2000, 2000, 3), dtype=np.uint8) * 255)
+    # for cls, *box in labels[:1000]:
+    #     ImageDraw.Draw(img).rectangle(box, width=1, outline=colors[int(cls) % 10])  # plot
+    ax[1].imshow(img)
+    ax[1].axis('off')
+
+    for a in [0, 1, 2, 3]:
+        for s in ['top', 'right', 'left', 'bottom']:
+            ax[a].spines[s].set_visible(False)
+
+    plt.savefig(save_dir / 'labels.jpg', dpi=200)
+    matplotlib.use('Agg')
+    plt.close()
+
+    # loggers
+    for k, v in loggers.items() or {}:
+        if k == 'wandb' and v:
+            v.log({"Labels": [v.Image(str(x), caption=x.name) for x in save_dir.glob('*labels*.jpg')]})
+
+
+def plot_evolution(yaml_file='data/hyp.finetune.yaml'):  # from utils.plots import *; plot_evolution()
+    # Plot hyperparameter evolution results in evolve.txt
+    with open(yaml_file) as f:
+        hyp = yaml.load(f, Loader=yaml.SafeLoader)
+    x = np.loadtxt('evolve.txt', ndmin=2)
+    f = fitness(x)
+    # weights = (f - f.min()) ** 2  # for weighted results
+    plt.figure(figsize=(10, 12), tight_layout=True)
+    matplotlib.rc('font', **{'size': 8})
+    for i, (k, v) in enumerate(hyp.items()):
+        y = x[:, i + 7]
+        # mu = (y * weights).sum() / weights.sum()  # best weighted result
+        mu = y[f.argmax()]  # best single result
+        plt.subplot(6, 5, i + 1)
+        plt.scatter(y, f, c=hist2d(y, f, 20), cmap='viridis', alpha=.8, edgecolors='none')
+        plt.plot(mu, f.max(), 'k+', markersize=15)
+        plt.title('%s = %.3g' % (k, mu), fontdict={'size': 9})  # limit to 40 characters
+        if i % 5 != 0:
+            plt.yticks([])
+        print('%15s: %.3g' % (k, mu))
+    plt.savefig('evolve.png', dpi=200)
+    print('\nPlot saved as evolve.png')
+
+
+def profile_idetection(start=0, stop=0, labels=(), save_dir=''):
+    # Plot iDetection '*.txt' per-image logs. from utils.plots import *; profile_idetection()
+    ax = plt.subplots(2, 4, figsize=(12, 6), tight_layout=True)[1].ravel()
+    s = ['Images', 'Free Storage (GB)', 'RAM Usage (GB)', 'Battery', 'dt_raw (ms)', 'dt_smooth (ms)', 'real-world FPS']
+    files = list(Path(save_dir).glob('frames*.txt'))
+    for fi, f in enumerate(files):
+        try:
+            results = np.loadtxt(f, ndmin=2).T[:, 90:-30]  # clip first and last rows
+            n = results.shape[1]  # number of rows
+            x = np.arange(start, min(stop, n) if stop else n)
+            results = results[:, x]
+            t = (results[0] - results[0].min())  # set t0=0s
+            results[0] = x
+            for i, a in enumerate(ax):
+                if i < len(results):
+                    label = labels[fi] if len(labels) else f.stem.replace('frames_', '')
+                    a.plot(t, results[i], marker='.', label=label, linewidth=1, markersize=5)
+                    a.set_title(s[i])
+                    a.set_xlabel('time (s)')
+                    # if fi == len(files) - 1:
+                    #     a.set_ylim(bottom=0)
+                    for side in ['top', 'right']:
+                        a.spines[side].set_visible(False)
+                else:
+                    a.remove()
+        except Exception as e:
+            print('Warning: Plotting error for %s; %s' % (f, e))
+
+    ax[1].legend()
+    plt.savefig(Path(save_dir) / 'idetection_profile.png', dpi=200)
+
+
+def plot_results_overlay(start=0, stop=0):  # from utils.plots import *; plot_results_overlay()
+    # Plot training 'results*.txt', overlaying train and val losses
+    s = ['train', 'train', 'train', 'Precision', 'mAP@0.5', 'val', 'val', 'val', 'Recall', 'mAP@0.5:0.95']  # legends
+    t = ['Box', 'Objectness', 'Classification', 'P-R', 'mAP-F1']  # titles
+    for f in sorted(glob.glob('results*.txt') + glob.glob('../../Downloads/results*.txt')):
+        results = np.loadtxt(f, usecols=[2, 3, 4, 8, 9, 12, 13, 14, 10, 11], ndmin=2).T
+        n = results.shape[1]  # number of rows
+        x = range(start, min(stop, n) if stop else n)
+        fig, ax = plt.subplots(1, 5, figsize=(14, 3.5), tight_layout=True)
+        ax = ax.ravel()
+        for i in range(5):
+            for j in [i, i + 5]:
+                y = results[j, x]
+                ax[i].plot(x, y, marker='.', label=s[j])
+                # y_smooth = butter_lowpass_filtfilt(y)
+                # ax[i].plot(x, np.gradient(y_smooth), marker='.', label=s[j])
+
+            ax[i].set_title(t[i])
+            ax[i].legend()
+            ax[i].set_ylabel(f) if i == 0 else None  # add filename
+        fig.savefig(f.replace('.txt', '.png'), dpi=200)
+
+
+def plot_results(start=0, stop=0, bucket='', id=(), labels=(), save_dir=''):
+    # Plot training 'results*.txt'. from utils.plots import *; plot_results(save_dir='runs/train/exp')
+    fig, ax = plt.subplots(2, 5, figsize=(12, 6), tight_layout=True)
+    ax = ax.ravel()
+    s = ['Box', 'Objectness', 'Classification', 'Precision', 'Recall',
+         'val Box', 'val Objectness', 'val Classification', 'mAP@0.5', 'mAP@0.5:0.95']
+    if bucket:
+        # files = ['https://storage.googleapis.com/%s/results%g.txt' % (bucket, x) for x in id]
+        files = ['results%g.txt' % x for x in id]
+        c = ('gsutil cp ' + '%s ' * len(files) + '.') % tuple('gs://%s/results%g.txt' % (bucket, x) for x in id)
+        os.system(c)
+    else:
+        files = list(Path(save_dir).glob('results*.txt'))
+    assert len(files), 'No results.txt files found in %s, nothing to plot.' % os.path.abspath(save_dir)
+    for fi, f in enumerate(files):
+        try:
+            results = np.loadtxt(f, usecols=[2, 3, 4, 8, 9, 12, 13, 14, 10, 11], ndmin=2).T
+            n = results.shape[1]  # number of rows
+            x = range(start, min(stop, n) if stop else n)
+            for i in range(10):
+                y = results[i, x]
+                if i in [0, 1, 2, 5, 6, 7]:
+                    y[y == 0] = np.nan  # don't show zero loss values
+                    # y /= y[0]  # normalize
+                label = labels[fi] if len(labels) else f.stem
+                ax[i].plot(x, y, marker='.', label=label, linewidth=2, markersize=8)
+                ax[i].set_title(s[i])
+                # if i in [5, 6, 7]:  # share train and val loss y axes
+                #     ax[i].get_shared_y_axes().join(ax[i], ax[i - 5])
+        except Exception as e:
+            print('Warning: Plotting error for %s; %s' % (f, e))
+
+    ax[1].legend()
+    fig.savefig(Path(save_dir) / 'results.png', dpi=200)

utils/torch_utils.py

@@ -0,0 +1,294 @@
+# PyTorch utils
+
+import logging
+import math
+import os
+import subprocess
+import time
+from contextlib import contextmanager
+from copy import deepcopy
+from pathlib import Path
+
+import torch
+import torch.backends.cudnn as cudnn
+import torch.nn as nn
+import torch.nn.functional as F
+import torchvision
+
+try:
+    import thop  # for FLOPS computation
+except ImportError:
+    thop = None
+logger = logging.getLogger(__name__)
+
+
+@contextmanager
+def torch_distributed_zero_first(local_rank: int):
+    """
+    Decorator to make all processes in distributed training wait for each local_master to do something.
+    """
+    if local_rank not in [-1, 0]:
+        torch.distributed.barrier()
+    yield
+    if local_rank == 0:
+        torch.distributed.barrier()
+
+
+def init_torch_seeds(seed=0):
+    # Speed-reproducibility tradeoff https://pytorch.org/docs/stable/notes/randomness.html
+    torch.manual_seed(seed)
+    if seed == 0:  # slower, more reproducible
+        cudnn.benchmark, cudnn.deterministic = False, True
+    else:  # faster, less reproducible
+        cudnn.benchmark, cudnn.deterministic = True, False
+
+
+def git_describe():
+    # return human-readable git description, i.e. v5.0-5-g3e25f1e https://git-scm.com/docs/git-describe
+    if Path('.git').exists():
+        return subprocess.check_output('git describe --tags --long --always', shell=True).decode('utf-8')[:-1]
+    else:
+        return ''
+
+
+def select_device(device='', batch_size=None):
+    # device = 'cpu' or '0' or '0,1,2,3'
+    s = f'YOLOv5 {git_describe()} torch {torch.__version__} '  # string
+    cpu = device.lower() == 'cpu'
+    if cpu:
+        os.environ['CUDA_VISIBLE_DEVICES'] = '-1'  # force torch.cuda.is_available() = False
+    elif device:  # non-cpu device requested
+        os.environ['CUDA_VISIBLE_DEVICES'] = device  # set environment variable
+        assert torch.cuda.is_available(), f'CUDA unavailable, invalid device {device} requested'  # check availability
+
+    cuda = not cpu and torch.cuda.is_available()
+    if cuda:
+        n = torch.cuda.device_count()
+        if n > 1 and batch_size:  # check that batch_size is compatible with device_count
+            assert batch_size % n == 0, f'batch-size {batch_size} not multiple of GPU count {n}'
+        space = ' ' * len(s)
+        for i, d in enumerate(device.split(',') if device else range(n)):
+            p = torch.cuda.get_device_properties(i)
+            s += f"{'' if i == 0 else space}CUDA:{d} ({p.name}, {p.total_memory / 1024 ** 2}MB)\n"  # bytes to MB
+    else:
+        s += 'CPU\n'
+
+    logger.info(s)  # skip a line
+    return torch.device('cuda:0' if cuda else 'cpu')
+
+
+def time_synchronized():
+    # pytorch-accurate time
+    if torch.cuda.is_available():
+        torch.cuda.synchronize()
+    return time.time()
+
+
+def profile(x, ops, n=100, device=None):
+    # profile a pytorch module or list of modules. Example usage:
+    #     x = torch.randn(16, 3, 640, 640)  # input
+    #     m1 = lambda x: x * torch.sigmoid(x)
+    #     m2 = nn.SiLU()
+    #     profile(x, [m1, m2], n=100)  # profile speed over 100 iterations
+
+    device = device or torch.device('cuda:0' if torch.cuda.is_available() else 'cpu')
+    x = x.to(device)
+    x.requires_grad = True
+    print(torch.__version__, device.type, torch.cuda.get_device_properties(0) if device.type == 'cuda' else '')
+    print(f"\n{'Params':>12s}{'GFLOPS':>12s}{'forward (ms)':>16s}{'backward (ms)':>16s}{'input':>24s}{'output':>24s}")
+    for m in ops if isinstance(ops, list) else [ops]:
+        m = m.to(device) if hasattr(m, 'to') else m  # device
+        m = m.half() if hasattr(m, 'half') and isinstance(x, torch.Tensor) and x.dtype is torch.float16 else m  # type
+        dtf, dtb, t = 0., 0., [0., 0., 0.]  # dt forward, backward
+        try:
+            flops = thop.profile(m, inputs=(x,), verbose=False)[0] / 1E9 * 2  # GFLOPS
+        except:
+            flops = 0
+
+        for _ in range(n):
+            t[0] = time_synchronized()
+            y = m(x)
+            t[1] = time_synchronized()
+            try:
+                _ = y.sum().backward()
+                t[2] = time_synchronized()
+            except:  # no backward method
+                t[2] = float('nan')
+            dtf += (t[1] - t[0]) * 1000 / n  # ms per op forward
+            dtb += (t[2] - t[1]) * 1000 / n  # ms per op backward
+
+        s_in = tuple(x.shape) if isinstance(x, torch.Tensor) else 'list'
+        s_out = tuple(y.shape) if isinstance(y, torch.Tensor) else 'list'
+        p = sum(list(x.numel() for x in m.parameters())) if isinstance(m, nn.Module) else 0  # parameters
+        print(f'{p:12.4g}{flops:12.4g}{dtf:16.4g}{dtb:16.4g}{str(s_in):>24s}{str(s_out):>24s}')
+
+
+def is_parallel(model):
+    return type(model) in (nn.parallel.DataParallel, nn.parallel.DistributedDataParallel)
+
+
+def intersect_dicts(da, db, exclude=()):
+    # Dictionary intersection of matching keys and shapes, omitting 'exclude' keys, using da values
+    return {k: v for k, v in da.items() if k in db and not any(x in k for x in exclude) and v.shape == db[k].shape}
+
+
+def initialize_weights(model):
+    for m in model.modules():
+        t = type(m)
+        if t is nn.Conv2d:
+            pass  # nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
+        elif t is nn.BatchNorm2d:
+            m.eps = 1e-3
+            m.momentum = 0.03
+        elif t in [nn.Hardswish, nn.LeakyReLU, nn.ReLU, nn.ReLU6]:
+            m.inplace = True
+
+
+def find_modules(model, mclass=nn.Conv2d):
+    # Finds layer indices matching module class 'mclass'
+    return [i for i, m in enumerate(model.module_list) if isinstance(m, mclass)]
+
+
+def sparsity(model):
+    # Return global model sparsity
+    a, b = 0., 0.
+    for p in model.parameters():
+        a += p.numel()
+        b += (p == 0).sum()
+    return b / a
+
+
+def prune(model, amount=0.3):
+    # Prune model to requested global sparsity
+    import torch.nn.utils.prune as prune
+    print('Pruning model... ', end='')
+    for name, m in model.named_modules():
+        if isinstance(m, nn.Conv2d):
+            prune.l1_unstructured(m, name='weight', amount=amount)  # prune
+            prune.remove(m, 'weight')  # make permanent
+    print(' %.3g global sparsity' % sparsity(model))
+
+
+def fuse_conv_and_bn(conv, bn):
+    # Fuse convolution and batchnorm layers https://tehnokv.com/posts/fusing-batchnorm-and-conv/
+    fusedconv = nn.Conv2d(conv.in_channels,
+                          conv.out_channels,
+                          kernel_size=conv.kernel_size,
+                          stride=conv.stride,
+                          padding=conv.padding,
+                          groups=conv.groups,
+                          bias=True).requires_grad_(False).to(conv.weight.device)
+
+    # prepare filters
+    w_conv = conv.weight.clone().view(conv.out_channels, -1)
+    w_bn = torch.diag(bn.weight.div(torch.sqrt(bn.eps + bn.running_var)))
+    fusedconv.weight.copy_(torch.mm(w_bn, w_conv).view(fusedconv.weight.size()))
+
+    # prepare spatial bias
+    b_conv = torch.zeros(conv.weight.size(0), device=conv.weight.device) if conv.bias is None else conv.bias
+    b_bn = bn.bias - bn.weight.mul(bn.running_mean).div(torch.sqrt(bn.running_var + bn.eps))
+    fusedconv.bias.copy_(torch.mm(w_bn, b_conv.reshape(-1, 1)).reshape(-1) + b_bn)
+
+    return fusedconv
+
+
+def model_info(model, verbose=False, img_size=640):
+    # Model information. img_size may be int or list, i.e. img_size=640 or img_size=[640, 320]
+    n_p = sum(x.numel() for x in model.parameters())  # number parameters
+    n_g = sum(x.numel() for x in model.parameters() if x.requires_grad)  # number gradients
+    if verbose:
+        print('%5s %40s %9s %12s %20s %10s %10s' % ('layer', 'name', 'gradient', 'parameters', 'shape', 'mu', 'sigma'))
+        for i, (name, p) in enumerate(model.named_parameters()):
+            name = name.replace('module_list.', '')
+            print('%5g %40s %9s %12g %20s %10.3g %10.3g' %
+                  (i, name, p.requires_grad, p.numel(), list(p.shape), p.mean(), p.std()))
+
+    try:  # FLOPS
+        from thop import profile
+        stride = int(model.stride.max()) if hasattr(model, 'stride') else 32
+        img = torch.zeros((1, model.yaml.get('ch', 3), stride, stride), device=next(model.parameters()).device)  # input
+        flops = profile(deepcopy(model), inputs=(img,), verbose=False)[0] / 1E9 * 2  # stride GFLOPS
+        img_size = img_size if isinstance(img_size, list) else [img_size, img_size]  # expand if int/float
+        fs = ', %.1f GFLOPS' % (flops * img_size[0] / stride * img_size[1] / stride)  # 640x640 GFLOPS
+    except (ImportError, Exception):
+        fs = ''
+
+    logger.info(f"Model Summary: {len(list(model.modules()))} layers, {n_p} parameters, {n_g} gradients{fs}")
+
+
+def load_classifier(name='resnet101', n=2):
+    # Loads a pretrained model reshaped to n-class output
+    model = torchvision.models.__dict__[name](pretrained=True)
+
+    # ResNet model properties
+    # input_size = [3, 224, 224]
+    # input_space = 'RGB'
+    # input_range = [0, 1]
+    # mean = [0.485, 0.456, 0.406]
+    # std = [0.229, 0.224, 0.225]
+
+    # Reshape output to n classes
+    filters = model.fc.weight.shape[1]
+    model.fc.bias = nn.Parameter(torch.zeros(n), requires_grad=True)
+    model.fc.weight = nn.Parameter(torch.zeros(n, filters), requires_grad=True)
+    model.fc.out_features = n
+    return model
+
+
+def scale_img(img, ratio=1.0, same_shape=False, gs=32):  # img(16,3,256,416)
+    # scales img(bs,3,y,x) by ratio constrained to gs-multiple
+    if ratio == 1.0:
+        return img
+    else:
+        h, w = img.shape[2:]
+        s = (int(h * ratio), int(w * ratio))  # new size
+        img = F.interpolate(img, size=s, mode='bilinear', align_corners=False)  # resize
+        if not same_shape:  # pad/crop img
+            h, w = [math.ceil(x * ratio / gs) * gs for x in (h, w)]
+        return F.pad(img, [0, w - s[1], 0, h - s[0]], value=0.447)  # value = imagenet mean
+
+
+def copy_attr(a, b, include=(), exclude=()):
+    # Copy attributes from b to a, options to only include [...] and to exclude [...]
+    for k, v in b.__dict__.items():
+        if (len(include) and k not in include) or k.startswith('_') or k in exclude:
+            continue
+        else:
+            setattr(a, k, v)
+
+
+class ModelEMA:
+    """ Model Exponential Moving Average from https://github.com/rwightman/pytorch-image-models
+    Keep a moving average of everything in the model state_dict (parameters and buffers).
+    This is intended to allow functionality like
+    https://www.tensorflow.org/api_docs/python/tf/train/ExponentialMovingAverage
+    A smoothed version of the weights is necessary for some training schemes to perform well.
+    This class is sensitive where it is initialized in the sequence of model init,
+    GPU assignment and distributed training wrappers.
+    """
+
+    def __init__(self, model, decay=0.9999, updates=0):
+        # Create EMA
+        self.ema = deepcopy(model.module if is_parallel(model) else model).eval()  # FP32 EMA
+        # if next(model.parameters()).device.type != 'cpu':
+        #     self.ema.half()  # FP16 EMA
+        self.updates = updates  # number of EMA updates
+        self.decay = lambda x: decay * (1 - math.exp(-x / 2000))  # decay exponential ramp (to help early epochs)
+        for p in self.ema.parameters():
+            p.requires_grad_(False)
+
+    def update(self, model):
+        # Update EMA parameters
+        with torch.no_grad():
+            self.updates += 1
+            d = self.decay(self.updates)
+
+            msd = model.module.state_dict() if is_parallel(model) else model.state_dict()  # model state_dict
+            for k, v in self.ema.state_dict().items():
+                if v.dtype.is_floating_point:
+                    v *= d
+                    v += (1. - d) * msd[k].detach()
+
+    def update_attr(self, model, include=(), exclude=('process_group', 'reducer')):
+        # Update EMA attributes
+        copy_attr(self.ema, model, include, exclude)

utils/wandb_logging/log_dataset.py

@@ -0,0 +1,24 @@
+import argparse
+
+import yaml
+
+from wandb_utils import WandbLogger
+
+WANDB_ARTIFACT_PREFIX = 'wandb-artifact://'
+
+
+def create_dataset_artifact(opt):
+    with open(opt.data) as f:
+        data = yaml.load(f, Loader=yaml.SafeLoader)  # data dict
+    logger = WandbLogger(opt, '', None, data, job_type='Dataset Creation')
+
+
+if __name__ == '__main__':
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--data', type=str, default='data/coco128.yaml', help='data.yaml path')
+    parser.add_argument('--single-cls', action='store_true', help='train as single-class dataset')
+    parser.add_argument('--project', type=str, default='YOLOv5', help='name of W&B Project')
+    opt = parser.parse_args()
+    opt.resume = False  # Explicitly disallow resume check for dataset upload job
+
+    create_dataset_artifact(opt)

utils/wandb_logging/wandb_utils.py

@@ -0,0 +1,306 @@
+import json
+import sys
+from pathlib import Path
+
+import torch
+import yaml
+from tqdm import tqdm
+
+sys.path.append(str(Path(__file__).parent.parent.parent))  # add utils/ to path
+from utils.datasets import LoadImagesAndLabels
+from utils.datasets import img2label_paths
+from utils.general import colorstr, xywh2xyxy, check_dataset
+
+try:
+    import wandb
+    from wandb import init, finish
+except ImportError:
+    wandb = None
+
+WANDB_ARTIFACT_PREFIX = 'wandb-artifact://'
+
+
+def remove_prefix(from_string, prefix=WANDB_ARTIFACT_PREFIX):
+    return from_string[len(prefix):]
+
+
+def check_wandb_config_file(data_config_file):
+    wandb_config = '_wandb.'.join(data_config_file.rsplit('.', 1))  # updated data.yaml path
+    if Path(wandb_config).is_file():
+        return wandb_config
+    return data_config_file
+
+
+def get_run_info(run_path):
+    run_path = Path(remove_prefix(run_path, WANDB_ARTIFACT_PREFIX))
+    run_id = run_path.stem
+    project = run_path.parent.stem
+    model_artifact_name = 'run_' + run_id + '_model'
+    return run_id, project, model_artifact_name
+
+
+def check_wandb_resume(opt):
+    process_wandb_config_ddp_mode(opt) if opt.global_rank not in [-1, 0] else None
+    if isinstance(opt.resume, str):
+        if opt.resume.startswith(WANDB_ARTIFACT_PREFIX):
+            if opt.global_rank not in [-1, 0]:  # For resuming DDP runs
+                run_id, project, model_artifact_name = get_run_info(opt.resume)
+                api = wandb.Api()
+                artifact = api.artifact(project + '/' + model_artifact_name + ':latest')
+                modeldir = artifact.download()
+                opt.weights = str(Path(modeldir) / "last.pt")
+            return True
+    return None
+
+
+def process_wandb_config_ddp_mode(opt):
+    with open(opt.data) as f:
+        data_dict = yaml.load(f, Loader=yaml.SafeLoader)  # data dict
+    train_dir, val_dir = None, None
+    if isinstance(data_dict['train'], str) and data_dict['train'].startswith(WANDB_ARTIFACT_PREFIX):
+        api = wandb.Api()
+        train_artifact = api.artifact(remove_prefix(data_dict['train']) + ':' + opt.artifact_alias)
+        train_dir = train_artifact.download()
+        train_path = Path(train_dir) / 'data/images/'
+        data_dict['train'] = str(train_path)
+
+    if isinstance(data_dict['val'], str) and data_dict['val'].startswith(WANDB_ARTIFACT_PREFIX):
+        api = wandb.Api()
+        val_artifact = api.artifact(remove_prefix(data_dict['val']) + ':' + opt.artifact_alias)
+        val_dir = val_artifact.download()
+        val_path = Path(val_dir) / 'data/images/'
+        data_dict['val'] = str(val_path)
+    if train_dir or val_dir:
+        ddp_data_path = str(Path(val_dir) / 'wandb_local_data.yaml')
+        with open(ddp_data_path, 'w') as f:
+            yaml.dump(data_dict, f)
+        opt.data = ddp_data_path
+
+
+class WandbLogger():
+    def __init__(self, opt, name, run_id, data_dict, job_type='Training'):
+        # Pre-training routine --
+        self.job_type = job_type
+        self.wandb, self.wandb_run, self.data_dict = wandb, None if not wandb else wandb.run, data_dict
+        # It's more elegant to stick to 1 wandb.init call, but useful config data is overwritten in the WandbLogger's wandb.init call
+        if isinstance(opt.resume, str):  # checks resume from artifact
+            if opt.resume.startswith(WANDB_ARTIFACT_PREFIX):
+                run_id, project, model_artifact_name = get_run_info(opt.resume)
+                model_artifact_name = WANDB_ARTIFACT_PREFIX + model_artifact_name
+                assert wandb, 'install wandb to resume wandb runs'
+                # Resume wandb-artifact:// runs here| workaround for not overwriting wandb.config
+                self.wandb_run = wandb.init(id=run_id, project=project, resume='allow')
+                opt.resume = model_artifact_name
+        elif self.wandb:
+            self.wandb_run = wandb.init(config=opt,
+                                        resume="allow",
+                                        project='YOLOv5' if opt.project == 'runs/train' else Path(opt.project).stem,
+                                        name=name,
+                                        job_type=job_type,
+                                        id=run_id) if not wandb.run else wandb.run
+        if self.wandb_run:
+            if self.job_type == 'Training':
+                if not opt.resume:
+                    wandb_data_dict = self.check_and_upload_dataset(opt) if opt.upload_dataset else data_dict
+                    # Info useful for resuming from artifacts
+                    self.wandb_run.config.opt = vars(opt)
+                    self.wandb_run.config.data_dict = wandb_data_dict
+                self.data_dict = self.setup_training(opt, data_dict)
+            if self.job_type == 'Dataset Creation':
+                self.data_dict = self.check_and_upload_dataset(opt)
+        else:
+            prefix = colorstr('wandb: ')
+            print(f"{prefix}Install Weights & Biases for YOLOv5 logging with 'pip install wandb' (recommended)")
+
+    def check_and_upload_dataset(self, opt):
+        assert wandb, 'Install wandb to upload dataset'
+        check_dataset(self.data_dict)
+        config_path = self.log_dataset_artifact(opt.data,
+                                                opt.single_cls,
+                                                'YOLOv5' if opt.project == 'runs/train' else Path(opt.project).stem)
+        print("Created dataset config file ", config_path)
+        with open(config_path) as f:
+            wandb_data_dict = yaml.load(f, Loader=yaml.SafeLoader)
+        return wandb_data_dict
+
+    def setup_training(self, opt, data_dict):
+        self.log_dict, self.current_epoch, self.log_imgs = {}, 0, 16  # Logging Constants
+        self.bbox_interval = opt.bbox_interval
+        if isinstance(opt.resume, str):
+            modeldir, _ = self.download_model_artifact(opt)
+            if modeldir:
+                self.weights = Path(modeldir) / "last.pt"
+                config = self.wandb_run.config
+                opt.weights, opt.save_period, opt.batch_size, opt.bbox_interval, opt.epochs, opt.hyp = str(
+                    self.weights), config.save_period, config.total_batch_size, config.bbox_interval, config.epochs, \
+                                                                                                       config.opt['hyp']
+            data_dict = dict(self.wandb_run.config.data_dict)  # eliminates the need for config file to resume
+        if 'val_artifact' not in self.__dict__:  # If --upload_dataset is set, use the existing artifact, don't download
+            self.train_artifact_path, self.train_artifact = self.download_dataset_artifact(data_dict.get('train'),
+                                                                                           opt.artifact_alias)
+            self.val_artifact_path, self.val_artifact = self.download_dataset_artifact(data_dict.get('val'),
+                                                                                       opt.artifact_alias)
+            self.result_artifact, self.result_table, self.val_table, self.weights = None, None, None, None
+            if self.train_artifact_path is not None:
+                train_path = Path(self.train_artifact_path) / 'data/images/'
+                data_dict['train'] = str(train_path)
+            if self.val_artifact_path is not None:
+                val_path = Path(self.val_artifact_path) / 'data/images/'
+                data_dict['val'] = str(val_path)
+                self.val_table = self.val_artifact.get("val")
+                self.map_val_table_path()
+        if self.val_artifact is not None:
+            self.result_artifact = wandb.Artifact("run_" + wandb.run.id + "_progress", "evaluation")
+            self.result_table = wandb.Table(["epoch", "id", "prediction", "avg_confidence"])
+        if opt.bbox_interval == -1:
+            self.bbox_interval = opt.bbox_interval = (opt.epochs // 10) if opt.epochs > 10 else 1
+        return data_dict
+
+    def download_dataset_artifact(self, path, alias):
+        if isinstance(path, str) and path.startswith(WANDB_ARTIFACT_PREFIX):
+            dataset_artifact = wandb.use_artifact(remove_prefix(path, WANDB_ARTIFACT_PREFIX) + ":" + alias)
+            assert dataset_artifact is not None, "'Error: W&B dataset artifact doesn\'t exist'"
+            datadir = dataset_artifact.download()
+            return datadir, dataset_artifact
+        return None, None
+
+    def download_model_artifact(self, opt):
+        if opt.resume.startswith(WANDB_ARTIFACT_PREFIX):
+            model_artifact = wandb.use_artifact(remove_prefix(opt.resume, WANDB_ARTIFACT_PREFIX) + ":latest")
+            assert model_artifact is not None, 'Error: W&B model artifact doesn\'t exist'
+            modeldir = model_artifact.download()
+            epochs_trained = model_artifact.metadata.get('epochs_trained')
+            total_epochs = model_artifact.metadata.get('total_epochs')
+            assert epochs_trained < total_epochs, 'training to %g epochs is finished, nothing to resume.' % (
+                total_epochs)
+            return modeldir, model_artifact
+        return None, None
+
+    def log_model(self, path, opt, epoch, fitness_score, best_model=False):
+        model_artifact = wandb.Artifact('run_' + wandb.run.id + '_model', type='model', metadata={
+            'original_url': str(path),
+            'epochs_trained': epoch + 1,
+            'save period': opt.save_period,
+            'project': opt.project,
+            'total_epochs': opt.epochs,
+            'fitness_score': fitness_score
+        })
+        model_artifact.add_file(str(path / 'last.pt'), name='last.pt')
+        wandb.log_artifact(model_artifact,
+                           aliases=['latest', 'epoch ' + str(self.current_epoch), 'best' if best_model else ''])
+        print("Saving model artifact on epoch ", epoch + 1)
+
+    def log_dataset_artifact(self, data_file, single_cls, project, overwrite_config=False):
+        with open(data_file) as f:
+            data = yaml.load(f, Loader=yaml.SafeLoader)  # data dict
+        nc, names = (1, ['item']) if single_cls else (int(data['nc']), data['names'])
+        names = {k: v for k, v in enumerate(names)}  # to index dictionary
+        self.train_artifact = self.create_dataset_table(LoadImagesAndLabels(
+            data['train']), names, name='train') if data.get('train') else None
+        self.val_artifact = self.create_dataset_table(LoadImagesAndLabels(
+            data['val']), names, name='val') if data.get('val') else None
+        if data.get('train'):
+            data['train'] = WANDB_ARTIFACT_PREFIX + str(Path(project) / 'train')
+        if data.get('val'):
+            data['val'] = WANDB_ARTIFACT_PREFIX + str(Path(project) / 'val')
+        path = data_file if overwrite_config else '_wandb.'.join(data_file.rsplit('.', 1))  # updated data.yaml path
+        data.pop('download', None)
+        with open(path, 'w') as f:
+            yaml.dump(data, f)
+
+        if self.job_type == 'Training':  # builds correct artifact pipeline graph
+            self.wandb_run.use_artifact(self.val_artifact)
+            self.wandb_run.use_artifact(self.train_artifact)
+            self.val_artifact.wait()
+            self.val_table = self.val_artifact.get('val')
+            self.map_val_table_path()
+        else:
+            self.wandb_run.log_artifact(self.train_artifact)
+            self.wandb_run.log_artifact(self.val_artifact)
+        return path
+
+    def map_val_table_path(self):
+        self.val_table_map = {}
+        print("Mapping dataset")
+        for i, data in enumerate(tqdm(self.val_table.data)):
+            self.val_table_map[data[3]] = data[0]
+
+    def create_dataset_table(self, dataset, class_to_id, name='dataset'):
+        # TODO: Explore multiprocessing to slpit this loop parallely| This is essential for speeding up the the logging
+        artifact = wandb.Artifact(name=name, type="dataset")
+        img_files = tqdm([dataset.path]) if isinstance(dataset.path, str) and Path(dataset.path).is_dir() else None
+        img_files = tqdm(dataset.img_files) if not img_files else img_files
+        for img_file in img_files:
+            if Path(img_file).is_dir():
+                artifact.add_dir(img_file, name='data/images')
+                labels_path = 'labels'.join(dataset.path.rsplit('images', 1))
+                artifact.add_dir(labels_path, name='data/labels')
+            else:
+                artifact.add_file(img_file, name='data/images/' + Path(img_file).name)
+                label_file = Path(img2label_paths([img_file])[0])
+                artifact.add_file(str(label_file),
+                                  name='data/labels/' + label_file.name) if label_file.exists() else None
+        table = wandb.Table(columns=["id", "train_image", "Classes", "name"])
+        class_set = wandb.Classes([{'id': id, 'name': name} for id, name in class_to_id.items()])
+        for si, (img, labels, paths, shapes) in enumerate(tqdm(dataset)):
+            height, width = shapes[0]
+            labels[:, 2:] = (xywh2xyxy(labels[:, 2:].view(-1, 4))) * torch.Tensor([width, height, width, height])
+            box_data, img_classes = [], {}
+            for cls, *xyxy in labels[:, 1:].tolist():
+                cls = int(cls)
+                box_data.append({"position": {"minX": xyxy[0], "minY": xyxy[1], "maxX": xyxy[2], "maxY": xyxy[3]},
+                                 "class_id": cls,
+                                 "box_caption": "%s" % (class_to_id[cls]),
+                                 "scores": {"acc": 1},
+                                 "domain": "pixel"})
+                img_classes[cls] = class_to_id[cls]
+            boxes = {"ground_truth": {"box_data": box_data, "class_labels": class_to_id}}  # inference-space
+            table.add_data(si, wandb.Image(paths, classes=class_set, boxes=boxes), json.dumps(img_classes),
+                           Path(paths).name)
+        artifact.add(table, name)
+        return artifact
+
+    def log_training_progress(self, predn, path, names):
+        if self.val_table and self.result_table:
+            class_set = wandb.Classes([{'id': id, 'name': name} for id, name in names.items()])
+            box_data = []
+            total_conf = 0
+            for *xyxy, conf, cls in predn.tolist():
+                if conf >= 0.25:
+                    box_data.append(
+                        {"position": {"minX": xyxy[0], "minY": xyxy[1], "maxX": xyxy[2], "maxY": xyxy[3]},
+                         "class_id": int(cls),
+                         "box_caption": "%s %.3f" % (names[cls], conf),
+                         "scores": {"class_score": conf},
+                         "domain": "pixel"})
+                    total_conf = total_conf + conf
+            boxes = {"predictions": {"box_data": box_data, "class_labels": names}}  # inference-space
+            id = self.val_table_map[Path(path).name]
+            self.result_table.add_data(self.current_epoch,
+                                       id,
+                                       wandb.Image(self.val_table.data[id][1], boxes=boxes, classes=class_set),
+                                       total_conf / max(1, len(box_data))
+                                       )
+
+    def log(self, log_dict):
+        if self.wandb_run:
+            for key, value in log_dict.items():
+                self.log_dict[key] = value
+
+    def end_epoch(self, best_result=False):
+        if self.wandb_run:
+            wandb.log(self.log_dict)
+            self.log_dict = {}
+            if self.result_artifact:
+                train_results = wandb.JoinedTable(self.val_table, self.result_table, "id")
+                self.result_artifact.add(train_results, 'result')
+                wandb.log_artifact(self.result_artifact, aliases=['latest', 'epoch ' + str(self.current_epoch),
+                                                                  ('best' if best_result else '')])
+                self.result_table = wandb.Table(["epoch", "id", "prediction", "avg_confidence"])
+                self.result_artifact = wandb.Artifact("run_" + wandb.run.id + "_progress", "evaluation")
+
+    def finish_run(self):
+        if self.wandb_run:
+            if self.log_dict:
+                wandb.log(self.log_dict)
+            wandb.run.finish()

weights/download_weights.sh

@@ -0,0 +1,12 @@
+#!/bin/bash
+# Download latest models from https://github.com/ultralytics/yolov5/releases
+# Usage:
+#    $ bash weights/download_weights.sh
+
+python3 - <<EOF
+from utils.google_utils import attempt_download
+
+for x in ['s', 'm', 'l', 'x']:
+    attempt_download(f'yolov5{x}.pt')
+
+EOF

widerface_evaluate/evaluation.py

@@ -0,0 +1,303 @@
+"""
+WiderFace evaluation code
+author: wondervictor
+mail: tianhengcheng@gmail.com
+copyright@wondervictor
+"""
+
+import os
+import tqdm
+import pickle
+import argparse
+import numpy as np
+from scipy.io import loadmat
+from bbox import bbox_overlaps
+from IPython import embed
+
+
+def get_gt_boxes(gt_dir):
+    """ gt dir: (wider_face_val.mat, wider_easy_val.mat, wider_medium_val.mat, wider_hard_val.mat)"""
+
+    gt_mat = loadmat(os.path.join(gt_dir, 'wider_face_val.mat'))
+    hard_mat = loadmat(os.path.join(gt_dir, 'wider_hard_val.mat'))
+    medium_mat = loadmat(os.path.join(gt_dir, 'wider_medium_val.mat'))
+    easy_mat = loadmat(os.path.join(gt_dir, 'wider_easy_val.mat'))
+
+    facebox_list = gt_mat['face_bbx_list']
+    event_list = gt_mat['event_list']
+    file_list = gt_mat['file_list']
+
+    hard_gt_list = hard_mat['gt_list']
+    medium_gt_list = medium_mat['gt_list']
+    easy_gt_list = easy_mat['gt_list']
+
+    return facebox_list, event_list, file_list, hard_gt_list, medium_gt_list, easy_gt_list
+
+
+def get_gt_boxes_from_txt(gt_path, cache_dir):
+
+    cache_file = os.path.join(cache_dir, 'gt_cache.pkl')
+    if os.path.exists(cache_file):
+        f = open(cache_file, 'rb')
+        boxes = pickle.load(f)
+        f.close()
+        return boxes
+
+    f = open(gt_path, 'r')
+    state = 0
+    lines = f.readlines()
+    lines = list(map(lambda x: x.rstrip('\r\n'), lines))
+    boxes = {}
+    print(len(lines))
+    f.close()
+    current_boxes = []
+    current_name = None
+    for line in lines:
+        if state == 0 and '--' in line:
+            state = 1
+            current_name = line
+            continue
+        if state == 1:
+            state = 2
+            continue
+
+        if state == 2 and '--' in line:
+            state = 1
+            boxes[current_name] = np.array(current_boxes).astype('float32')
+            current_name = line
+            current_boxes = []
+            continue
+
+        if state == 2:
+            box = [float(x) for x in line.split(' ')[:4]]
+            current_boxes.append(box)
+            continue
+
+    f = open(cache_file, 'wb')
+    pickle.dump(boxes, f)
+    f.close()
+    return boxes
+
+
+def read_pred_file(filepath):
+
+    with open(filepath, 'r') as f:
+        lines = f.readlines()
+        img_file = lines[0].rstrip('\n\r')
+        lines = lines[2:]
+
+    # b = lines[0].rstrip('\r\n').split(' ')[:-1]
+    # c = float(b)
+    # a = map(lambda x: [[float(a[0]), float(a[1]), float(a[2]), float(a[3]), float(a[4])] for a in x.rstrip('\r\n').split(' ')], lines)
+    boxes = []
+    for line in lines:
+        line = line.rstrip('\r\n').split(' ')
+        if line[0] == '':
+            continue
+        # a = float(line[4])
+        boxes.append([float(line[0]), float(line[1]), float(line[2]), float(line[3]), float(line[4])])
+    boxes = np.array(boxes)
+    # boxes = np.array(list(map(lambda x: [float(a) for a in x.rstrip('\r\n').split(' ')], lines))).astype('float')
+    return img_file.split('/')[-1], boxes
+
+
+def get_preds(pred_dir):
+    events = os.listdir(pred_dir)
+    boxes = dict()
+    pbar = tqdm.tqdm(events)
+
+    for event in pbar:
+        pbar.set_description('Reading Predictions ')
+        event_dir = os.path.join(pred_dir, event)
+        event_images = os.listdir(event_dir)
+        current_event = dict()
+        for imgtxt in event_images:
+            imgname, _boxes = read_pred_file(os.path.join(event_dir, imgtxt))
+            current_event[imgname.rstrip('.jpg')] = _boxes
+        boxes[event] = current_event
+    return boxes
+
+
+def norm_score(pred):
+    """ norm score
+    pred {key: [[x1,y1,x2,y2,s]]}
+    """
+
+    max_score = 0
+    min_score = 1
+
+    for _, k in pred.items():
+        for _, v in k.items():
+            if len(v) == 0:
+                continue
+            _min = np.min(v[:, -1])
+            _max = np.max(v[:, -1])
+            max_score = max(_max, max_score)
+            min_score = min(_min, min_score)
+
+    diff = max_score - min_score
+    for _, k in pred.items():
+        for _, v in k.items():
+            if len(v) == 0:
+                continue
+            v[:, -1] = (v[:, -1] - min_score)/diff
+
+
+def image_eval(pred, gt, ignore, iou_thresh):
+    """ single image evaluation
+    pred: Nx5
+    gt: Nx4
+    ignore:
+    """
+
+    _pred = pred.copy()
+    _gt = gt.copy()
+    pred_recall = np.zeros(_pred.shape[0])
+    recall_list = np.zeros(_gt.shape[0])
+    proposal_list = np.ones(_pred.shape[0])
+
+    _pred[:, 2] = _pred[:, 2] + _pred[:, 0]
+    _pred[:, 3] = _pred[:, 3] + _pred[:, 1]
+    _gt[:, 2] = _gt[:, 2] + _gt[:, 0]
+    _gt[:, 3] = _gt[:, 3] + _gt[:, 1]
+
+    overlaps = bbox_overlaps(_pred[:, :4], _gt)
+
+    for h in range(_pred.shape[0]):
+
+        gt_overlap = overlaps[h]
+        max_overlap, max_idx = gt_overlap.max(), gt_overlap.argmax()
+        if max_overlap >= iou_thresh:
+            if ignore[max_idx] == 0:
+                recall_list[max_idx] = -1
+                proposal_list[h] = -1
+            elif recall_list[max_idx] == 0:
+                recall_list[max_idx] = 1
+
+        r_keep_index = np.where(recall_list == 1)[0]
+        pred_recall[h] = len(r_keep_index)
+    return pred_recall, proposal_list
+
+
+def img_pr_info(thresh_num, pred_info, proposal_list, pred_recall):
+    pr_info = np.zeros((thresh_num, 2)).astype('float')
+    for t in range(thresh_num):
+
+        thresh = 1 - (t+1)/thresh_num
+        r_index = np.where(pred_info[:, 4] >= thresh)[0]
+        if len(r_index) == 0:
+            pr_info[t, 0] = 0
+            pr_info[t, 1] = 0
+        else:
+            r_index = r_index[-1]
+            p_index = np.where(proposal_list[:r_index+1] == 1)[0]
+            pr_info[t, 0] = len(p_index)
+            pr_info[t, 1] = pred_recall[r_index]
+    return pr_info
+
+
+def dataset_pr_info(thresh_num, pr_curve, count_face):
+    _pr_curve = np.zeros((thresh_num, 2))
+    for i in range(thresh_num):
+        _pr_curve[i, 0] = pr_curve[i, 1] / pr_curve[i, 0]
+        _pr_curve[i, 1] = pr_curve[i, 1] / count_face
+    return _pr_curve
+
+
+def voc_ap(rec, prec):
+
+    # correct AP calculation
+    # first append sentinel values at the end
+    mrec = np.concatenate(([0.], rec, [1.]))
+    mpre = np.concatenate(([0.], prec, [0.]))
+
+    # compute the precision envelope
+    for i in range(mpre.size - 1, 0, -1):
+        mpre[i - 1] = np.maximum(mpre[i - 1], mpre[i])
+
+    # to calculate area under PR curve, look for points
+    # where X axis (recall) changes value
+    i = np.where(mrec[1:] != mrec[:-1])[0]
+
+    # and sum (\Delta recall) * prec
+    ap = np.sum((mrec[i + 1] - mrec[i]) * mpre[i + 1])
+    return ap
+
+
+def evaluation(pred, gt_path, iou_thresh=0.5):
+    pred = get_preds(pred)
+    norm_score(pred)
+    facebox_list, event_list, file_list, hard_gt_list, medium_gt_list, easy_gt_list = get_gt_boxes(gt_path)
+    event_num = len(event_list)
+    thresh_num = 1000
+    settings = ['easy', 'medium', 'hard']
+    setting_gts = [easy_gt_list, medium_gt_list, hard_gt_list]
+    aps = []
+    for setting_id in range(3):
+        # different setting
+        gt_list = setting_gts[setting_id]
+        count_face = 0
+        pr_curve = np.zeros((thresh_num, 2)).astype('float')
+        # [hard, medium, easy]
+        pbar = tqdm.tqdm(range(event_num))
+        for i in pbar:
+            pbar.set_description('Processing {}'.format(settings[setting_id]))
+            event_name = str(event_list[i][0][0])
+            img_list = file_list[i][0]
+            pred_list = pred[event_name]
+            sub_gt_list = gt_list[i][0]
+            # img_pr_info_list = np.zeros((len(img_list), thresh_num, 2))
+            gt_bbx_list = facebox_list[i][0]
+
+            for j in range(len(img_list)):
+                pred_info = pred_list[str(img_list[j][0][0])]
+
+                gt_boxes = gt_bbx_list[j][0].astype('float')
+                keep_index = sub_gt_list[j][0]
+                count_face += len(keep_index)
+
+                if len(gt_boxes) == 0 or len(pred_info) == 0:
+                    continue
+                ignore = np.zeros(gt_boxes.shape[0])
+                if len(keep_index) != 0:
+                    ignore[keep_index-1] = 1
+                pred_recall, proposal_list = image_eval(pred_info, gt_boxes, ignore, iou_thresh)
+
+                _img_pr_info = img_pr_info(thresh_num, pred_info, proposal_list, pred_recall)
+
+                pr_curve += _img_pr_info
+        pr_curve = dataset_pr_info(thresh_num, pr_curve, count_face)
+
+        propose = pr_curve[:, 0]
+        recall = pr_curve[:, 1]
+
+        ap = voc_ap(recall, propose)
+        aps.append(ap)
+
+    print("==================== Results ====================")
+    print("Easy   Val AP: {}".format(aps[0]))
+    print("Medium Val AP: {}".format(aps[1]))
+    print("Hard   Val AP: {}".format(aps[2]))
+    print("=================================================")
+
+
+if __name__ == '__main__':
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument('-p', '--pred', default="./widerface_txt/")
+    parser.add_argument('-g', '--gt', default='./ground_truth/')
+
+    args = parser.parse_args()
+    evaluation(args.pred, args.gt)
+
+
+
+
+
+
+
+
+
+
+
+

widerface_evaluate/setup.py

@@ -0,0 +1,13 @@
+"""
+WiderFace evaluation code
+author: wondervictor
+mail: tianhengcheng@gmail.com
+copyright@wondervictor
+"""
+
+from distutils.core import setup, Extension
+from Cython.Build import cythonize
+import numpy
+
+package = Extension('bbox', ['box_overlaps.pyx'], include_dirs=[numpy.get_include()])
+setup(ext_modules=cythonize([package]))