From f91a6126eee74f5cf918b6fe57455160529926ea Mon Sep 17 00:00:00 2001 From: felixhjh <852142024@qq.com> Date: Tue, 29 Nov 2022 08:26:27 +0000 Subject: [PATCH 01/30] Update keypointdetection result docs --- docs/api/vision_results/keypointdetection_result.md | 8 ++------ docs/api_docs/python/vision_results_cn.md | 3 +-- docs/api_docs/python/vision_results_en.md | 3 +-- tests/models/test_ppyoloe.py | 10 +++++++--- 4 files changed, 11 insertions(+), 13 deletions(-) diff --git a/docs/api/vision_results/keypointdetection_result.md b/docs/api/vision_results/keypointdetection_result.md index a47057ad8..645880d23 100644 --- a/docs/api/vision_results/keypointdetection_result.md +++ b/docs/api/vision_results/keypointdetection_result.md @@ -16,16 +16,13 @@ struct KeyPointDetectionResult { }; ``` -- **keypoints**: 成员变量，表示识别到的目标行为的关键点坐标。`keypoints.size()= N * J * 2`， +- **keypoints**: 成员变量，表示识别到的目标行为的关键点坐标。`keypoints.size()= N * J`， - `N`：图片中的目标数量 - `J`：num_joints（一个目标的关键点数量） - - `3`:坐标信息[x, y] - **scores**: 成员变量，表示识别到的目标行为的关键点坐标的置信度。`scores.size()= N * J` - `N`：图片中的目标数量 - `J`:num_joints（一个目标的关键点数量） - **num_joints**: 成员变量，一个目标的关键点数量 - -- **num_joints**: 成员变量，一个目标的关键点数量 - **Clear()**: 成员函数，用于清除结构体中存储的结果 - **Str()**: 成员函数，将结构体中的信息以字符串形式输出（用于Debug） @@ -34,10 +31,9 @@ struct KeyPointDetectionResult { `fastdeploy.vision.KeyPointDetectionResult` - **keypoints**(list of list(float)): 成员变量，表示识别到的目标行为的关键点坐标。 - `keypoints.size()= N * J * 2` + `keypoints.size()= N * J` `N`:图片中的目标数量 `J`:num_joints（关键点数量） - `3`:坐标信息[x, y, conf] - **scores**(list of float): 成员变量，表示识别到的目标行为的关键点坐标的置信度。 `scores.size()= N * J` `N`:图片中的目标数量 diff --git a/docs/api_docs/python/vision_results_cn.md b/docs/api_docs/python/vision_results_cn.md index 19b2a6662..e9dc00621 100644 --- a/docs/api_docs/python/vision_results_cn.md +++ b/docs/api_docs/python/vision_results_cn.md @@ -46,10 +46,9 @@ API:`fastdeploy.vision.FaceDetectionResult` , 该结果返回: KeyPointDetectionResult 代码定义在`fastdeploy/vision/common/result.h`中，用于表明图像中目标行为的各个关键点坐标和置信度。 API:`fastdeploy.vision.KeyPointDetectionResult` , 该结果返回: -- **keypoints**(list of list(float)): 成员变量，表示识别到的目标行为的关键点坐标。`keypoints.size()= N * J * 2`， +- **keypoints**(list of list(float)): 成员变量，表示识别到的目标行为的关键点坐标。`keypoints.size()= N * J`， - `N`：图片中的目标数量 - `J`：num_joints（一个目标的关键点数量） - - `3`:坐标信息[x, y] - **scores**(list of float): 成员变量，表示识别到的目标行为的关键点坐标的置信度。`scores.size()= N * J` - `N`：图片中的目标数量 - `J`:num_joints（一个目标的关键点数量） diff --git a/docs/api_docs/python/vision_results_en.md b/docs/api_docs/python/vision_results_en.md index dc02d0b4f..f297331d1 100644 --- a/docs/api_docs/python/vision_results_en.md +++ b/docs/api_docs/python/vision_results_en.md @@ -49,10 +49,9 @@ API: `fastdeploy.vision.FaceDetectionResult`, The FaceDetectionResult will retur The KeyPointDetectionResult code is defined in `fastdeploy/vision/common/result.h` and is used to indicate the coordinates and confidence of each keypoint of the target behavior in the image. API:`fastdeploy.vision.KeyPointDetectionResult`, The KeyPointDetectionResult will return: -- **keypoints**(list of list(float)): Member variable, representing the key point coordinates of the identified target behavior. `keypoints.size()= N * J * 2`， +- **keypoints**(list of list(float)): Member variable, representing the key point coordinates of the identified target behavior. `keypoints.size()= N * J`， - `N`: number of objects in the picture - `J`: num_joints（number of keypoints for a target） - - `3`: 坐标信息[x, y] - **scores**(list of float): Member variable, representing the confidence of the keypoint coordinates of the recognized target behavior. `scores.size()= N * J` - `N`: number of objects in the picture - `J`: num_joints（number of keypoints for a target） diff --git a/tests/models/test_ppyoloe.py b/tests/models/test_ppyoloe.py index 08b19bf91..a479f90e1 100755 --- a/tests/models/test_ppyoloe.py +++ b/tests/models/test_ppyoloe.py @@ -60,6 +60,7 @@ def test_detection_ppyoloe(): assert diff_label_ids[scores > score_threshold].max( ) < 1e-04, "There's diff in label_ids." + def test_detection_ppyoloe1(): model_url = "https://bj.bcebos.com/paddlehub/fastdeploy/ppyoloe_crn_l_300e_coco.tgz" input_url1 = "https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg" @@ -75,15 +76,18 @@ def test_detection_ppyoloe1(): preprocessor = fd.vision.detection.PaddleDetPreprocessor(config_file) postprocessor = fd.vision.detection.PaddleDetPostprocessor() - + rc.test_option.set_model_path(model_file, params_file) - runtime = fd.Runtime(rc.test_option); + runtime = fd.Runtime(rc.test_option) # compare diff im1 = cv2.imread("./resources/000000014439.jpg") for i in range(2): input_tensors = preprocessor.run([im1]) - output_tensors = runtime.infer({"image": input_tensors[0], "scale_factor": input_tensors[1]}) + output_tensors = runtime.infer({ + "image": input_tensors[0], + "scale_factor": input_tensors[1] + }) results = postprocessor.run(output_tensors) result = results[0] with open("resources/ppyoloe_baseline.pkl", "rb") as f: From 2f6f6977d76fff2176e46c796ad3551c5b911c7b Mon Sep 17 00:00:00 2001 From: felixhjh <852142024@qq.com> Date: Sun, 11 Dec 2022 14:10:16 +0000 Subject: [PATCH 02/30] Update im.copy() to im in examples --- examples/vision/classification/resnet/python/infer.py | 2 +- examples/vision/classification/yolov5cls/python/infer.py | 2 +- examples/vision/detection/nanodet_plus/python/infer.py | 2 +- examples/vision/detection/scaledyolov4/python/infer.py | 2 +- examples/vision/detection/yolor/python/infer.py | 2 +- examples/vision/detection/yolov5/python/infer.py | 2 +- examples/vision/detection/yolov5/quantize/python/infer.py | 2 +- examples/vision/detection/yolov5lite/python/infer.py | 2 +- examples/vision/detection/yolov6/python/infer.py | 2 +- examples/vision/detection/yolov6/quantize/python/infer.py | 2 +- examples/vision/detection/yolov7/python/infer.py | 2 +- examples/vision/detection/yolov7/quantize/python/infer.py | 2 +- examples/vision/detection/yolov7end2end_ort/python/infer.py | 2 +- examples/vision/detection/yolov7end2end_trt/python/infer.py | 2 +- examples/vision/detection/yolox/python/infer.py | 2 +- examples/vision/facealign/face_landmark_1000/python/infer.py | 2 +- examples/vision/facealign/pfld/python/infer.py | 2 +- examples/vision/facealign/pipnet/python/infer.py | 2 +- examples/vision/facedet/retinaface/python/infer.py | 2 +- examples/vision/facedet/scrfd/python/infer.py | 2 +- examples/vision/facedet/scrfd/rknpu2/python/infer.py | 2 +- examples/vision/facedet/ultraface/python/infer.py | 2 +- examples/vision/facedet/yolov5face/python/infer.py | 2 +- examples/vision/headpose/fsanet/python/infer.py | 2 +- examples/vision/matting/modnet/python/infer.py | 2 +- examples/vision/matting/ppmatting/python/infer.py | 2 +- examples/vision/matting/rvm/python/infer.py | 4 ++-- examples/vision/segmentation/paddleseg/python/infer.py | 2 +- .../vision/segmentation/paddleseg/quantize/python/infer.py | 2 +- examples/vision/segmentation/paddleseg/rknpu2/python/infer.py | 2 +- 30 files changed, 31 insertions(+), 31 deletions(-) diff --git a/examples/vision/classification/resnet/python/infer.py b/examples/vision/classification/resnet/python/infer.py index b8b268f3a..ba22304fc 100644 --- a/examples/vision/classification/resnet/python/infer.py +++ b/examples/vision/classification/resnet/python/infer.py @@ -46,5 +46,5 @@ model = fd.vision.classification.ResNet( args.model, runtime_option=runtime_option) # 预测图片分类结果 im = cv2.imread(args.image) -result = model.predict(im.copy(), args.topk) +result = model.predict(im, args.topk) print(result) diff --git a/examples/vision/classification/yolov5cls/python/infer.py b/examples/vision/classification/yolov5cls/python/infer.py index 576db32f2..55974a764 100644 --- a/examples/vision/classification/yolov5cls/python/infer.py +++ b/examples/vision/classification/yolov5cls/python/infer.py @@ -47,5 +47,5 @@ model = fd.vision.classification.YOLOv5Cls( # 预测图片分类结果 im = cv2.imread(args.image) -result = model.predict(im.copy(), args.topk) +result = model.predict(im, args.topk) print(result) diff --git a/examples/vision/detection/nanodet_plus/python/infer.py b/examples/vision/detection/nanodet_plus/python/infer.py index 9e7bf184f..a60f3a3b4 100644 --- a/examples/vision/detection/nanodet_plus/python/infer.py +++ b/examples/vision/detection/nanodet_plus/python/infer.py @@ -52,7 +52,7 @@ if args.image is None: else: image = args.image im = cv2.imread(image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 diff --git a/examples/vision/detection/scaledyolov4/python/infer.py b/examples/vision/detection/scaledyolov4/python/infer.py index 9d990d77b..cd23e84e3 100644 --- a/examples/vision/detection/scaledyolov4/python/infer.py +++ b/examples/vision/detection/scaledyolov4/python/infer.py @@ -52,7 +52,7 @@ if args.image is None: else: image = args.image im = cv2.imread(image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 diff --git a/examples/vision/detection/yolor/python/infer.py b/examples/vision/detection/yolor/python/infer.py index d5ab3dd7f..4111fa0ae 100644 --- a/examples/vision/detection/yolor/python/infer.py +++ b/examples/vision/detection/yolor/python/infer.py @@ -54,7 +54,7 @@ else: image = args.image im = cv2.imread(image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 diff --git a/examples/vision/detection/yolov5/python/infer.py b/examples/vision/detection/yolov5/python/infer.py index fdded06c9..462740e9c 100644 --- a/examples/vision/detection/yolov5/python/infer.py +++ b/examples/vision/detection/yolov5/python/infer.py @@ -52,7 +52,7 @@ if args.image is None: else: image = args.image im = cv2.imread(image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 diff --git a/examples/vision/detection/yolov5/quantize/python/infer.py b/examples/vision/detection/yolov5/quantize/python/infer.py index da502fe93..996bc5419 100644 --- a/examples/vision/detection/yolov5/quantize/python/infer.py +++ b/examples/vision/detection/yolov5/quantize/python/infer.py @@ -71,7 +71,7 @@ model = fd.vision.detection.YOLOv5( # 预测图片检测结果 im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 diff --git a/examples/vision/detection/yolov5lite/python/infer.py b/examples/vision/detection/yolov5lite/python/infer.py index b1ec69046..2242a33a5 100644 --- a/examples/vision/detection/yolov5lite/python/infer.py +++ b/examples/vision/detection/yolov5lite/python/infer.py @@ -52,7 +52,7 @@ if args.image is None: else: image = args.image im = cv2.imread(image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 diff --git a/examples/vision/detection/yolov6/python/infer.py b/examples/vision/detection/yolov6/python/infer.py index 0a0a163fb..47bf3e689 100644 --- a/examples/vision/detection/yolov6/python/infer.py +++ b/examples/vision/detection/yolov6/python/infer.py @@ -52,7 +52,7 @@ if args.image is None: else: image = args.image im = cv2.imread(image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 diff --git a/examples/vision/detection/yolov6/quantize/python/infer.py b/examples/vision/detection/yolov6/quantize/python/infer.py index da9fa3d97..77f46d4c2 100644 --- a/examples/vision/detection/yolov6/quantize/python/infer.py +++ b/examples/vision/detection/yolov6/quantize/python/infer.py @@ -71,7 +71,7 @@ model = fd.vision.detection.YOLOv6( # 预测图片检测结果 im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 diff --git a/examples/vision/detection/yolov7/python/infer.py b/examples/vision/detection/yolov7/python/infer.py index b0ece7e47..468b47dc5 100644 --- a/examples/vision/detection/yolov7/python/infer.py +++ b/examples/vision/detection/yolov7/python/infer.py @@ -52,7 +52,7 @@ if args.image is None: else: image = args.image im = cv2.imread(image) -result = model.predict(im.copy()) +result = model.predict(im) # 预测结果可视化 vis_im = fd.vision.vis_detection(im, result) diff --git a/examples/vision/detection/yolov7/quantize/python/infer.py b/examples/vision/detection/yolov7/quantize/python/infer.py index de84e4061..d07834c7c 100644 --- a/examples/vision/detection/yolov7/quantize/python/infer.py +++ b/examples/vision/detection/yolov7/quantize/python/infer.py @@ -71,7 +71,7 @@ model = fd.vision.detection.YOLOv7( # 预测图片检测结果 im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 diff --git a/examples/vision/detection/yolov7end2end_ort/python/infer.py b/examples/vision/detection/yolov7end2end_ort/python/infer.py index 2b812b71a..914facb59 100644 --- a/examples/vision/detection/yolov7end2end_ort/python/infer.py +++ b/examples/vision/detection/yolov7end2end_ort/python/infer.py @@ -44,7 +44,7 @@ model = fd.vision.detection.YOLOv7End2EndORT( # 预测图片检测结果 im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 diff --git a/examples/vision/detection/yolov7end2end_trt/python/infer.py b/examples/vision/detection/yolov7end2end_trt/python/infer.py index d179de06f..b1ad8ea0c 100644 --- a/examples/vision/detection/yolov7end2end_trt/python/infer.py +++ b/examples/vision/detection/yolov7end2end_trt/python/infer.py @@ -44,7 +44,7 @@ model = fd.vision.detection.YOLOv7End2EndTRT( # 预测图片检测结果 im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 diff --git a/examples/vision/detection/yolox/python/infer.py b/examples/vision/detection/yolox/python/infer.py index 69203010b..c5e71754a 100644 --- a/examples/vision/detection/yolox/python/infer.py +++ b/examples/vision/detection/yolox/python/infer.py @@ -52,7 +52,7 @@ if args.image is None: else: image = args.image im = cv2.imread(image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 vis_im = fd.vision.vis_detection(im, result) diff --git a/examples/vision/facealign/face_landmark_1000/python/infer.py b/examples/vision/facealign/face_landmark_1000/python/infer.py index bbb3a19c8..9eaedfb63 100644 --- a/examples/vision/facealign/face_landmark_1000/python/infer.py +++ b/examples/vision/facealign/face_landmark_1000/python/infer.py @@ -82,7 +82,7 @@ model = fd.vision.facealign.FaceLandmark1000( # for image im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 可视化结果 vis_im = fd.vision.vis_face_alignment(im, result) diff --git a/examples/vision/facealign/pfld/python/infer.py b/examples/vision/facealign/pfld/python/infer.py index 622fbf822..0a695e8c6 100755 --- a/examples/vision/facealign/pfld/python/infer.py +++ b/examples/vision/facealign/pfld/python/infer.py @@ -80,7 +80,7 @@ model = fd.vision.facealign.PFLD(args.model, runtime_option=runtime_option) # for image im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 可视化结果 vis_im = fd.vision.vis_face_alignment(im, result) diff --git a/examples/vision/facealign/pipnet/python/infer.py b/examples/vision/facealign/pipnet/python/infer.py index 628b86f79..46f9c2d90 100644 --- a/examples/vision/facealign/pipnet/python/infer.py +++ b/examples/vision/facealign/pipnet/python/infer.py @@ -85,7 +85,7 @@ model = fd.vision.facealign.PIPNet(args.model, runtime_option=runtime_option) model.num_landmarks = args.num_landmarks # for image im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 可视化结果 vis_im = fd.vision.vis_face_alignment(im, result) diff --git a/examples/vision/facedet/retinaface/python/infer.py b/examples/vision/facedet/retinaface/python/infer.py index 3b0152b1c..6b7f63e5e 100644 --- a/examples/vision/facedet/retinaface/python/infer.py +++ b/examples/vision/facedet/retinaface/python/infer.py @@ -43,7 +43,7 @@ model = fd.vision.facedet.RetinaFace(args.model, runtime_option=runtime_option) # 预测图片检测结果 im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 vis_im = fd.vision.vis_face_detection(im, result) diff --git a/examples/vision/facedet/scrfd/python/infer.py b/examples/vision/facedet/scrfd/python/infer.py index a99e66385..af141e011 100644 --- a/examples/vision/facedet/scrfd/python/infer.py +++ b/examples/vision/facedet/scrfd/python/infer.py @@ -43,7 +43,7 @@ model = fd.vision.facedet.SCRFD(args.model, runtime_option=runtime_option) # 预测图片检测结果 im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 vis_im = fd.vision.vis_face_detection(im, result) diff --git a/examples/vision/facedet/scrfd/rknpu2/python/infer.py b/examples/vision/facedet/scrfd/rknpu2/python/infer.py index 3b3fc9d83..77e57a0f1 100644 --- a/examples/vision/facedet/scrfd/rknpu2/python/infer.py +++ b/examples/vision/facedet/scrfd/rknpu2/python/infer.py @@ -49,7 +49,7 @@ model.disable_normalize_and_permute() # 预测图片分割结果 im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 可视化结果 diff --git a/examples/vision/facedet/ultraface/python/infer.py b/examples/vision/facedet/ultraface/python/infer.py index 5399110b9..2087ce5bf 100644 --- a/examples/vision/facedet/ultraface/python/infer.py +++ b/examples/vision/facedet/ultraface/python/infer.py @@ -43,7 +43,7 @@ model = fd.vision.facedet.UltraFace(args.model, runtime_option=runtime_option) # 预测图片检测结果 im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 vis_im = fd.vision.vis_face_detection(im, result) diff --git a/examples/vision/facedet/yolov5face/python/infer.py b/examples/vision/facedet/yolov5face/python/infer.py index a9f044682..ca907afb4 100644 --- a/examples/vision/facedet/yolov5face/python/infer.py +++ b/examples/vision/facedet/yolov5face/python/infer.py @@ -43,7 +43,7 @@ model = fd.vision.facedet.YOLOv5Face(args.model, runtime_option=runtime_option) # 预测图片检测结果 im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 预测结果可视化 vis_im = fd.vision.vis_face_detection(im, result) diff --git a/examples/vision/headpose/fsanet/python/infer.py b/examples/vision/headpose/fsanet/python/infer.py index 866ce6d5c..488e35153 100644 --- a/examples/vision/headpose/fsanet/python/infer.py +++ b/examples/vision/headpose/fsanet/python/infer.py @@ -80,7 +80,7 @@ model = fd.vision.headpose.FSANet(args.model, runtime_option=runtime_option) # for image im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 可视化结果 vis_im = fd.vision.vis_headpose(im, result) diff --git a/examples/vision/matting/modnet/python/infer.py b/examples/vision/matting/modnet/python/infer.py index 408ba2340..abb9be037 100644 --- a/examples/vision/matting/modnet/python/infer.py +++ b/examples/vision/matting/modnet/python/infer.py @@ -52,7 +52,7 @@ model.size = (256, 256) # 预测图片抠图结果 im = cv2.imread(args.image) bg = cv2.imread(args.bg) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 可视化结果 vis_im = fd.vision.vis_matting_alpha(im, result) diff --git a/examples/vision/matting/ppmatting/python/infer.py b/examples/vision/matting/ppmatting/python/infer.py index 61031e1b6..32a81eca7 100644 --- a/examples/vision/matting/ppmatting/python/infer.py +++ b/examples/vision/matting/ppmatting/python/infer.py @@ -56,7 +56,7 @@ model = fd.vision.matting.PPMatting( # 预测图片抠图结果 im = cv2.imread(args.image) bg = cv2.imread(args.bg) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 可视化结果 vis_im = fd.vision.vis_matting(im, result) diff --git a/examples/vision/matting/rvm/python/infer.py b/examples/vision/matting/rvm/python/infer.py index 11951b00f..fcde64fb2 100755 --- a/examples/vision/matting/rvm/python/infer.py +++ b/examples/vision/matting/rvm/python/infer.py @@ -73,7 +73,7 @@ if args.video is not None: # for video cap = cv2.VideoCapture(args.video) # Define the codec and create VideoWriter object - fourcc = cv2.VideoWriter_fourcc(*'mp4v') + fourcc = cv2.VideoWriter_fourcc('m', 'p', '4', 'v') composition = cv2.VideoWriter(output_composition, fourcc, 20.0, (1080, 1920)) alpha = cv2.VideoWriter(output_alpha, fourcc, 20.0, (1080, 1920)) @@ -100,7 +100,7 @@ if args.video is not None: if args.image is not None: # for image im = cv2.imread(args.image) - result = model.predict(im.copy()) + result = model.predict(im) print(result) # 可视化结果 vis_im = fd.vision.vis_matting(im, result) diff --git a/examples/vision/segmentation/paddleseg/python/infer.py b/examples/vision/segmentation/paddleseg/python/infer.py index 866e32bfb..9df7665a2 100644 --- a/examples/vision/segmentation/paddleseg/python/infer.py +++ b/examples/vision/segmentation/paddleseg/python/infer.py @@ -49,7 +49,7 @@ model = fd.vision.segmentation.PaddleSegModel( # 预测图片分割结果 im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 可视化结果 diff --git a/examples/vision/segmentation/paddleseg/quantize/python/infer.py b/examples/vision/segmentation/paddleseg/quantize/python/infer.py index f95f04d17..85a875c1e 100644 --- a/examples/vision/segmentation/paddleseg/quantize/python/infer.py +++ b/examples/vision/segmentation/paddleseg/quantize/python/infer.py @@ -72,5 +72,5 @@ model = fd.vision.segmentation.PaddleSegModel( # 预测图片检测结果 im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) diff --git a/examples/vision/segmentation/paddleseg/rknpu2/python/infer.py b/examples/vision/segmentation/paddleseg/rknpu2/python/infer.py index 8841132a9..d7239eb42 100644 --- a/examples/vision/segmentation/paddleseg/rknpu2/python/infer.py +++ b/examples/vision/segmentation/paddleseg/rknpu2/python/infer.py @@ -53,7 +53,7 @@ model.disable_normalize_and_permute() # 预测图片分割结果 im = cv2.imread(args.image) -result = model.predict(im.copy()) +result = model.predict(im) print(result) # 可视化结果 From bfc9ac9ee1f0810fdbeb50761a96aa6a21b60285 Mon Sep 17 00:00:00 2001 From: felixhjh <852142024@qq.com> Date: Mon, 12 Dec 2022 04:13:40 +0000 Subject: [PATCH 03/30] Update new Api, fastdeploy::vision::Visualize to fastdeploy::vision --- examples/vision/detection/nanodet_plus/cpp/infer.cc | 6 +++--- examples/vision/detection/scaledyolov4/cpp/infer.cc | 6 +++--- examples/vision/detection/yolor/cpp/infer.cc | 6 +++--- examples/vision/detection/yolov5/cpp/infer.cc | 6 +++--- examples/vision/detection/yolov5/quantize/cpp/infer.cc | 2 +- examples/vision/detection/yolov5lite/cpp/infer.cc | 6 +++--- examples/vision/detection/yolov6/cpp/infer.cc | 6 +++--- examples/vision/detection/yolov6/quantize/cpp/infer.cc | 2 +- examples/vision/detection/yolov7/cpp/infer.cc | 6 +++--- examples/vision/detection/yolov7/quantize/cpp/infer.cc | 2 +- examples/vision/detection/yolov7end2end_ort/cpp/infer.cc | 6 +++--- examples/vision/detection/yolov7end2end_trt/cpp/infer.cc | 6 +++--- examples/vision/detection/yolox/cpp/infer.cc | 6 +++--- examples/vision/facedet/retinaface/cpp/infer.cc | 6 +++--- examples/vision/facedet/scrfd/cpp/infer.cc | 6 +++--- examples/vision/facedet/scrfd/rknpu2/cpp/infer.cc | 2 +- examples/vision/facedet/ultraface/cpp/infer.cc | 6 +++--- examples/vision/facedet/yolov5face/cpp/infer.cc | 6 +++--- examples/vision/matting/modnet/cpp/infer.cc | 6 +++--- examples/vision/matting/ppmatting/cpp/infer.cc | 6 +++--- 20 files changed, 52 insertions(+), 52 deletions(-) diff --git a/examples/vision/detection/nanodet_plus/cpp/infer.cc b/examples/vision/detection/nanodet_plus/cpp/infer.cc index 8443639cc..8c9f3338f 100644 --- a/examples/vision/detection/nanodet_plus/cpp/infer.cc +++ b/examples/vision/detection/nanodet_plus/cpp/infer.cc @@ -30,7 +30,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { return; } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -55,7 +55,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -82,7 +82,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/scaledyolov4/cpp/infer.cc b/examples/vision/detection/scaledyolov4/cpp/infer.cc index 7d912b223..040823a1d 100644 --- a/examples/vision/detection/scaledyolov4/cpp/infer.cc +++ b/examples/vision/detection/scaledyolov4/cpp/infer.cc @@ -31,7 +31,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -56,7 +56,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -83,7 +83,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolor/cpp/infer.cc b/examples/vision/detection/yolor/cpp/infer.cc index 0fe8913d4..105855a02 100644 --- a/examples/vision/detection/yolor/cpp/infer.cc +++ b/examples/vision/detection/yolor/cpp/infer.cc @@ -31,7 +31,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -55,7 +55,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -81,7 +81,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov5/cpp/infer.cc b/examples/vision/detection/yolov5/cpp/infer.cc index a7ac1fe81..dea29450b 100644 --- a/examples/vision/detection/yolov5/cpp/infer.cc +++ b/examples/vision/detection/yolov5/cpp/infer.cc @@ -31,7 +31,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -55,7 +55,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -81,7 +81,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov5/quantize/cpp/infer.cc b/examples/vision/detection/yolov5/quantize/cpp/infer.cc index e429b2aad..690c19ddf 100644 --- a/examples/vision/detection/yolov5/quantize/cpp/infer.cc +++ b/examples/vision/detection/yolov5/quantize/cpp/infer.cc @@ -39,7 +39,7 @@ void InitAndInfer(const std::string& model_dir, const std::string& image_file, std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov5lite/cpp/infer.cc b/examples/vision/detection/yolov5lite/cpp/infer.cc index ac32bca93..0d9c58201 100644 --- a/examples/vision/detection/yolov5lite/cpp/infer.cc +++ b/examples/vision/detection/yolov5lite/cpp/infer.cc @@ -31,7 +31,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -56,7 +56,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -83,7 +83,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov6/cpp/infer.cc b/examples/vision/detection/yolov6/cpp/infer.cc index 72b2e7bed..fd00796f1 100644 --- a/examples/vision/detection/yolov6/cpp/infer.cc +++ b/examples/vision/detection/yolov6/cpp/infer.cc @@ -31,7 +31,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -55,7 +55,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -81,7 +81,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov6/quantize/cpp/infer.cc b/examples/vision/detection/yolov6/quantize/cpp/infer.cc index b40200962..57754ca30 100644 --- a/examples/vision/detection/yolov6/quantize/cpp/infer.cc +++ b/examples/vision/detection/yolov6/quantize/cpp/infer.cc @@ -39,7 +39,7 @@ void InitAndInfer(const std::string& model_dir, const std::string& image_file, std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov7/cpp/infer.cc b/examples/vision/detection/yolov7/cpp/infer.cc index cf79a16ad..803a7227a 100644 --- a/examples/vision/detection/yolov7/cpp/infer.cc +++ b/examples/vision/detection/yolov7/cpp/infer.cc @@ -31,7 +31,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -55,7 +55,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -81,7 +81,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov7/quantize/cpp/infer.cc b/examples/vision/detection/yolov7/quantize/cpp/infer.cc index 0eda80b6c..6033e3454 100644 --- a/examples/vision/detection/yolov7/quantize/cpp/infer.cc +++ b/examples/vision/detection/yolov7/quantize/cpp/infer.cc @@ -39,7 +39,7 @@ void InitAndInfer(const std::string& model_dir, const std::string& image_file, std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov7end2end_ort/cpp/infer.cc b/examples/vision/detection/yolov7end2end_ort/cpp/infer.cc index a0e70544a..72ffe1a91 100644 --- a/examples/vision/detection/yolov7end2end_ort/cpp/infer.cc +++ b/examples/vision/detection/yolov7end2end_ort/cpp/infer.cc @@ -31,7 +31,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -56,7 +56,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -83,7 +83,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov7end2end_trt/cpp/infer.cc b/examples/vision/detection/yolov7end2end_trt/cpp/infer.cc index 1c7a17d37..20d963edf 100644 --- a/examples/vision/detection/yolov7end2end_trt/cpp/infer.cc +++ b/examples/vision/detection/yolov7end2end_trt/cpp/infer.cc @@ -31,7 +31,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -56,7 +56,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -83,7 +83,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolox/cpp/infer.cc b/examples/vision/detection/yolox/cpp/infer.cc index 2eeaccbf8..bed65f982 100644 --- a/examples/vision/detection/yolox/cpp/infer.cc +++ b/examples/vision/detection/yolox/cpp/infer.cc @@ -31,7 +31,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -55,7 +55,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -81,7 +81,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/facedet/retinaface/cpp/infer.cc b/examples/vision/facedet/retinaface/cpp/infer.cc index a1fd27b6e..ceacce746 100644 --- a/examples/vision/facedet/retinaface/cpp/infer.cc +++ b/examples/vision/facedet/retinaface/cpp/infer.cc @@ -31,7 +31,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -55,7 +55,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -81,7 +81,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/facedet/scrfd/cpp/infer.cc b/examples/vision/facedet/scrfd/cpp/infer.cc index c804218ee..28823242a 100644 --- a/examples/vision/facedet/scrfd/cpp/infer.cc +++ b/examples/vision/facedet/scrfd/cpp/infer.cc @@ -31,7 +31,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -55,7 +55,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -81,7 +81,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/facedet/scrfd/rknpu2/cpp/infer.cc b/examples/vision/facedet/scrfd/rknpu2/cpp/infer.cc index a01f1b184..cd06004fe 100644 --- a/examples/vision/facedet/scrfd/rknpu2/cpp/infer.cc +++ b/examples/vision/facedet/scrfd/rknpu2/cpp/infer.cc @@ -73,7 +73,7 @@ void InferScrfd(const std::string& device) { (dur / CLOCKS_PER_SEC)); std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisFaceDetection(im, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } \ No newline at end of file diff --git a/examples/vision/facedet/ultraface/cpp/infer.cc b/examples/vision/facedet/ultraface/cpp/infer.cc index b45bb3b0b..6327c63d6 100644 --- a/examples/vision/facedet/ultraface/cpp/infer.cc +++ b/examples/vision/facedet/ultraface/cpp/infer.cc @@ -31,7 +31,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -55,7 +55,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -81,7 +81,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/facedet/yolov5face/cpp/infer.cc b/examples/vision/facedet/yolov5face/cpp/infer.cc index 418834e1e..58e15c8ef 100644 --- a/examples/vision/facedet/yolov5face/cpp/infer.cc +++ b/examples/vision/facedet/yolov5face/cpp/infer.cc @@ -31,7 +31,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -55,7 +55,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -81,7 +81,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::Visualize::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/matting/modnet/cpp/infer.cc b/examples/vision/matting/modnet/cpp/infer.cc index fe1ebc910..d11c9717f 100644 --- a/examples/vision/matting/modnet/cpp/infer.cc +++ b/examples/vision/matting/modnet/cpp/infer.cc @@ -34,7 +34,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file, auto vis_im = fastdeploy::vision::VisMatting(im_bak, res); auto vis_im_with_bg = - fastdeploy::vision::Visualize::SwapBackgroundMatting(im_bak, bg, res); + fastdeploy::vision::SwapBackgroundMatting(im_bak, bg, res); cv::imwrite("visualized_result.jpg", vis_im_with_bg); cv::imwrite("visualized_result_fg.jpg", vis_im); std::cout << "Visualized result save in ./visualized_result_replaced_bg.jpg " @@ -65,7 +65,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file, auto vis_im = fastdeploy::vision::VisMatting(im_bak, res); auto vis_im_with_bg = - fastdeploy::vision::Visualize::SwapBackgroundMatting(im_bak, bg, res); + fastdeploy::vision::SwapBackgroundMatting(im_bak, bg, res); cv::imwrite("visualized_result.jpg", vis_im_with_bg); cv::imwrite("visualized_result_fg.jpg", vis_im); std::cout << "Visualized result save in ./visualized_result_replaced_bg.jpg " @@ -97,7 +97,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file, auto vis_im = fastdeploy::vision::VisMatting(im_bak, res); auto vis_im_with_bg = - fastdeploy::vision::Visualize::SwapBackgroundMatting(im_bak, bg, res); + fastdeploy::vision::SwapBackgroundMatting(im_bak, bg, res); cv::imwrite("visualized_result.jpg", vis_im_with_bg); cv::imwrite("visualized_result_fg.jpg", vis_im); std::cout << "Visualized result save in ./visualized_result_replaced_bg.jpg " diff --git a/examples/vision/matting/ppmatting/cpp/infer.cc b/examples/vision/matting/ppmatting/cpp/infer.cc index 304e4239a..9c917a77f 100644 --- a/examples/vision/matting/ppmatting/cpp/infer.cc +++ b/examples/vision/matting/ppmatting/cpp/infer.cc @@ -44,7 +44,7 @@ void CpuInfer(const std::string& model_dir, const std::string& image_file, } auto vis_im = fastdeploy::vision::VisMatting(im_bak, res); auto vis_im_with_bg = - fastdeploy::vision::Visualize::SwapBackgroundMatting(im_bak, bg, res); + fastdeploy::vision::SwapBackgroundMatting(im_bak, bg, res); cv::imwrite("visualized_result.jpg", vis_im_with_bg); cv::imwrite("visualized_result_fg.jpg", vis_im); std::cout << "Visualized result save in ./visualized_result_replaced_bg.jpg " @@ -78,7 +78,7 @@ void GpuInfer(const std::string& model_dir, const std::string& image_file, } auto vis_im = fastdeploy::vision::VisMatting(im_bak, res); auto vis_im_with_bg = - fastdeploy::vision::Visualize::SwapBackgroundMatting(im_bak, bg, res); + fastdeploy::vision::SwapBackgroundMatting(im_bak, bg, res); cv::imwrite("visualized_result.jpg", vis_im_with_bg); cv::imwrite("visualized_result_fg.jpg", vis_im); std::cout << "Visualized result save in ./visualized_result_replaced_bg.jpg " @@ -113,7 +113,7 @@ void TrtInfer(const std::string& model_dir, const std::string& image_file, } auto vis_im = fastdeploy::vision::VisMatting(im_bak, res); auto vis_im_with_bg = - fastdeploy::vision::Visualize::SwapBackgroundMatting(im_bak, bg, res); + fastdeploy::vision::SwapBackgroundMatting(im_bak, bg, res); cv::imwrite("visualized_result.jpg", vis_im_with_bg); cv::imwrite("visualized_result_fg.jpg", vis_im); std::cout << "Visualized result save in ./visualized_result_replaced_bg.jpg " From 37359c65f3af287f97462ee8853786304fb52496 Mon Sep 17 00:00:00 2001 From: felixhjh <852142024@qq.com> Date: Mon, 12 Dec 2022 10:59:15 +0000 Subject: [PATCH 04/30] Update SwapBackgroundSegmentation && SwapBackgroundMatting to SwapBackground --- .../vision/detection/nanodet_plus/cpp/infer.cc | 9 +++------ .../vision/detection/scaledyolov4/cpp/infer.cc | 9 +++------ examples/vision/detection/yolor/cpp/infer.cc | 9 +++------ examples/vision/detection/yolov5/cpp/infer.cc | 9 +++------ .../vision/detection/yolov5/quantize/cpp/infer.cc | 3 +-- examples/vision/detection/yolov5lite/cpp/infer.cc | 9 +++------ examples/vision/detection/yolov6/cpp/infer.cc | 9 +++------ .../vision/detection/yolov6/quantize/cpp/infer.cc | 3 +-- examples/vision/detection/yolov7/cpp/infer.cc | 9 +++------ .../vision/detection/yolov7/quantize/cpp/infer.cc | 3 +-- .../detection/yolov7end2end_ort/cpp/infer.cc | 9 +++------ .../detection/yolov7end2end_trt/cpp/infer.cc | 11 ++++------- examples/vision/detection/yolox/cpp/infer.cc | 9 +++------ examples/vision/facedet/retinaface/cpp/infer.cc | 9 +++------ examples/vision/facedet/scrfd/cpp/infer.cc | 9 +++------ examples/vision/facedet/ultraface/cpp/infer.cc | 9 +++------ examples/vision/facedet/yolov5face/cpp/infer.cc | 9 +++------ examples/vision/matting/modnet/cpp/infer.cc | 15 ++++++--------- examples/vision/matting/modnet/python/infer.py | 2 +- examples/vision/matting/ppmatting/cpp/infer.cc | 15 ++++++--------- examples/vision/matting/ppmatting/python/infer.py | 2 +- examples/vision/matting/rvm/python/infer.py | 4 ++-- 22 files changed, 62 insertions(+), 113 deletions(-) diff --git a/examples/vision/detection/nanodet_plus/cpp/infer.cc b/examples/vision/detection/nanodet_plus/cpp/infer.cc index 8c9f3338f..ec0bbee98 100644 --- a/examples/vision/detection/nanodet_plus/cpp/infer.cc +++ b/examples/vision/detection/nanodet_plus/cpp/infer.cc @@ -22,7 +22,6 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -30,7 +29,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { return; } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -46,7 +45,6 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -55,7 +53,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -73,7 +71,6 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -82,7 +79,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/scaledyolov4/cpp/infer.cc b/examples/vision/detection/scaledyolov4/cpp/infer.cc index 040823a1d..683156cc9 100644 --- a/examples/vision/detection/scaledyolov4/cpp/infer.cc +++ b/examples/vision/detection/scaledyolov4/cpp/infer.cc @@ -22,7 +22,6 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -31,7 +30,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -47,7 +46,6 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -56,7 +54,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -74,7 +72,6 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -83,7 +80,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolor/cpp/infer.cc b/examples/vision/detection/yolor/cpp/infer.cc index 105855a02..bacc18b40 100644 --- a/examples/vision/detection/yolor/cpp/infer.cc +++ b/examples/vision/detection/yolor/cpp/infer.cc @@ -22,7 +22,6 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -31,7 +30,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -46,7 +45,6 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -55,7 +53,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -72,7 +70,6 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -81,7 +78,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov5/cpp/infer.cc b/examples/vision/detection/yolov5/cpp/infer.cc index dea29450b..1c3907918 100644 --- a/examples/vision/detection/yolov5/cpp/infer.cc +++ b/examples/vision/detection/yolov5/cpp/infer.cc @@ -22,7 +22,6 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -31,7 +30,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -46,7 +45,6 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -55,7 +53,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -72,7 +70,6 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -81,7 +78,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov5/quantize/cpp/infer.cc b/examples/vision/detection/yolov5/quantize/cpp/infer.cc index 690c19ddf..1addf1507 100644 --- a/examples/vision/detection/yolov5/quantize/cpp/infer.cc +++ b/examples/vision/detection/yolov5/quantize/cpp/infer.cc @@ -29,7 +29,6 @@ void InitAndInfer(const std::string& model_dir, const std::string& image_file, assert(model.Initialized()); auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -39,7 +38,7 @@ void InitAndInfer(const std::string& model_dir, const std::string& image_file, std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov5lite/cpp/infer.cc b/examples/vision/detection/yolov5lite/cpp/infer.cc index 0d9c58201..0a3f7b81b 100644 --- a/examples/vision/detection/yolov5lite/cpp/infer.cc +++ b/examples/vision/detection/yolov5lite/cpp/infer.cc @@ -22,7 +22,6 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -31,7 +30,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -47,7 +46,6 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -56,7 +54,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -74,7 +72,6 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -83,7 +80,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov6/cpp/infer.cc b/examples/vision/detection/yolov6/cpp/infer.cc index fd00796f1..dbca64f40 100644 --- a/examples/vision/detection/yolov6/cpp/infer.cc +++ b/examples/vision/detection/yolov6/cpp/infer.cc @@ -22,7 +22,6 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -31,7 +30,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -46,7 +45,6 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -55,7 +53,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -72,7 +70,6 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -81,7 +78,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov6/quantize/cpp/infer.cc b/examples/vision/detection/yolov6/quantize/cpp/infer.cc index 57754ca30..adda977f3 100644 --- a/examples/vision/detection/yolov6/quantize/cpp/infer.cc +++ b/examples/vision/detection/yolov6/quantize/cpp/infer.cc @@ -29,7 +29,6 @@ void InitAndInfer(const std::string& model_dir, const std::string& image_file, assert(model.Initialized()); auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -39,7 +38,7 @@ void InitAndInfer(const std::string& model_dir, const std::string& image_file, std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov7/cpp/infer.cc b/examples/vision/detection/yolov7/cpp/infer.cc index 803a7227a..5fd848b8e 100644 --- a/examples/vision/detection/yolov7/cpp/infer.cc +++ b/examples/vision/detection/yolov7/cpp/infer.cc @@ -22,7 +22,6 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -31,7 +30,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -46,7 +45,6 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -55,7 +53,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -72,7 +70,6 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -81,7 +78,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov7/quantize/cpp/infer.cc b/examples/vision/detection/yolov7/quantize/cpp/infer.cc index 6033e3454..9ae42b582 100644 --- a/examples/vision/detection/yolov7/quantize/cpp/infer.cc +++ b/examples/vision/detection/yolov7/quantize/cpp/infer.cc @@ -29,7 +29,6 @@ void InitAndInfer(const std::string& model_dir, const std::string& image_file, assert(model.Initialized()); auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -39,7 +38,7 @@ void InitAndInfer(const std::string& model_dir, const std::string& image_file, std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov7end2end_ort/cpp/infer.cc b/examples/vision/detection/yolov7end2end_ort/cpp/infer.cc index 72ffe1a91..5c9df2816 100644 --- a/examples/vision/detection/yolov7end2end_ort/cpp/infer.cc +++ b/examples/vision/detection/yolov7end2end_ort/cpp/infer.cc @@ -22,7 +22,6 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -31,7 +30,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -47,7 +46,6 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -56,7 +54,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -74,7 +72,6 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -83,7 +80,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/detection/yolov7end2end_trt/cpp/infer.cc b/examples/vision/detection/yolov7end2end_trt/cpp/infer.cc index 20d963edf..4f7a2f791 100644 --- a/examples/vision/detection/yolov7end2end_trt/cpp/infer.cc +++ b/examples/vision/detection/yolov7end2end_trt/cpp/infer.cc @@ -22,7 +22,6 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -31,7 +30,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -47,7 +46,6 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -56,7 +54,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -74,7 +72,6 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -83,7 +80,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -107,4 +104,4 @@ int main(int argc, char* argv[]) { TrtInfer(argv[1], argv[2]); } return 0; -} \ No newline at end of file +} diff --git a/examples/vision/detection/yolox/cpp/infer.cc b/examples/vision/detection/yolox/cpp/infer.cc index bed65f982..836ab7e63 100644 --- a/examples/vision/detection/yolox/cpp/infer.cc +++ b/examples/vision/detection/yolox/cpp/infer.cc @@ -22,7 +22,6 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -31,7 +30,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -46,7 +45,6 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -55,7 +53,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -72,7 +70,6 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::DetectionResult res; if (!model.Predict(&im, &res)) { @@ -81,7 +78,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/facedet/retinaface/cpp/infer.cc b/examples/vision/facedet/retinaface/cpp/infer.cc index ceacce746..f125e4ba7 100644 --- a/examples/vision/facedet/retinaface/cpp/infer.cc +++ b/examples/vision/facedet/retinaface/cpp/infer.cc @@ -22,7 +22,6 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::FaceDetectionResult res; if (!model.Predict(&im, &res)) { @@ -31,7 +30,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -46,7 +45,6 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::FaceDetectionResult res; if (!model.Predict(&im, &res)) { @@ -55,7 +53,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -72,7 +70,6 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::FaceDetectionResult res; if (!model.Predict(&im, &res)) { @@ -81,7 +78,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/facedet/scrfd/cpp/infer.cc b/examples/vision/facedet/scrfd/cpp/infer.cc index 28823242a..c09403741 100644 --- a/examples/vision/facedet/scrfd/cpp/infer.cc +++ b/examples/vision/facedet/scrfd/cpp/infer.cc @@ -22,7 +22,6 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::FaceDetectionResult res; if (!model.Predict(&im, &res)) { @@ -31,7 +30,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -46,7 +45,6 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::FaceDetectionResult res; if (!model.Predict(&im, &res)) { @@ -55,7 +53,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -72,7 +70,6 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::FaceDetectionResult res; if (!model.Predict(&im, &res)) { @@ -81,7 +78,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/facedet/ultraface/cpp/infer.cc b/examples/vision/facedet/ultraface/cpp/infer.cc index 6327c63d6..b50674e1d 100644 --- a/examples/vision/facedet/ultraface/cpp/infer.cc +++ b/examples/vision/facedet/ultraface/cpp/infer.cc @@ -22,7 +22,6 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::FaceDetectionResult res; if (!model.Predict(&im, &res)) { @@ -31,7 +30,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -46,7 +45,6 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::FaceDetectionResult res; if (!model.Predict(&im, &res)) { @@ -55,7 +53,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -72,7 +70,6 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::FaceDetectionResult res; if (!model.Predict(&im, &res)) { @@ -81,7 +78,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/facedet/yolov5face/cpp/infer.cc b/examples/vision/facedet/yolov5face/cpp/infer.cc index 58e15c8ef..ecdcfcd0c 100644 --- a/examples/vision/facedet/yolov5face/cpp/infer.cc +++ b/examples/vision/facedet/yolov5face/cpp/infer.cc @@ -22,7 +22,6 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::FaceDetectionResult res; if (!model.Predict(&im, &res)) { @@ -31,7 +30,7 @@ void CpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -46,7 +45,6 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::FaceDetectionResult res; if (!model.Predict(&im, &res)) { @@ -55,7 +53,7 @@ void GpuInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } @@ -72,7 +70,6 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } auto im = cv::imread(image_file); - auto im_bak = im.clone(); fastdeploy::vision::FaceDetectionResult res; if (!model.Predict(&im, &res)) { @@ -81,7 +78,7 @@ void TrtInfer(const std::string& model_file, const std::string& image_file) { } std::cout << res.Str() << std::endl; - auto vis_im = fastdeploy::vision::VisFaceDetection(im_bak, res); + auto vis_im = fastdeploy::vision::VisFaceDetection(im, res); cv::imwrite("vis_result.jpg", vis_im); std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; } diff --git a/examples/vision/matting/modnet/cpp/infer.cc b/examples/vision/matting/modnet/cpp/infer.cc index d11c9717f..aa280eddd 100644 --- a/examples/vision/matting/modnet/cpp/infer.cc +++ b/examples/vision/matting/modnet/cpp/infer.cc @@ -23,7 +23,6 @@ void CpuInfer(const std::string& model_file, const std::string& image_file, } model.size = {256, 256}; auto im = cv::imread(image_file); - auto im_bak = im.clone(); cv::Mat bg = cv::imread(background_file); fastdeploy::vision::MattingResult res; @@ -32,9 +31,9 @@ void CpuInfer(const std::string& model_file, const std::string& image_file, return; } - auto vis_im = fastdeploy::vision::VisMatting(im_bak, res); + auto vis_im = fastdeploy::vision::VisMatting(im, res); auto vis_im_with_bg = - fastdeploy::vision::SwapBackgroundMatting(im_bak, bg, res); + fastdeploy::vision::SwapBackground(im, bg, res); cv::imwrite("visualized_result.jpg", vis_im_with_bg); cv::imwrite("visualized_result_fg.jpg", vis_im); std::cout << "Visualized result save in ./visualized_result_replaced_bg.jpg " @@ -54,7 +53,6 @@ void GpuInfer(const std::string& model_file, const std::string& image_file, model.size = {256, 256}; auto im = cv::imread(image_file); - auto im_bak = im.clone(); cv::Mat bg = cv::imread(background_file); fastdeploy::vision::MattingResult res; @@ -63,9 +61,9 @@ void GpuInfer(const std::string& model_file, const std::string& image_file, return; } - auto vis_im = fastdeploy::vision::VisMatting(im_bak, res); + auto vis_im = fastdeploy::vision::VisMatting(im, res); auto vis_im_with_bg = - fastdeploy::vision::SwapBackgroundMatting(im_bak, bg, res); + fastdeploy::vision::SwapBackground(im, bg, res); cv::imwrite("visualized_result.jpg", vis_im_with_bg); cv::imwrite("visualized_result_fg.jpg", vis_im); std::cout << "Visualized result save in ./visualized_result_replaced_bg.jpg " @@ -86,7 +84,6 @@ void TrtInfer(const std::string& model_file, const std::string& image_file, } model.size = {256, 256}; auto im = cv::imread(image_file); - auto im_bak = im.clone(); cv::Mat bg = cv::imread(background_file); fastdeploy::vision::MattingResult res; @@ -95,9 +92,9 @@ void TrtInfer(const std::string& model_file, const std::string& image_file, return; } - auto vis_im = fastdeploy::vision::VisMatting(im_bak, res); + auto vis_im = fastdeploy::vision::VisMatting(im, res); auto vis_im_with_bg = - fastdeploy::vision::SwapBackgroundMatting(im_bak, bg, res); + fastdeploy::vision::SwapBackground(im, bg, res); cv::imwrite("visualized_result.jpg", vis_im_with_bg); cv::imwrite("visualized_result_fg.jpg", vis_im); std::cout << "Visualized result save in ./visualized_result_replaced_bg.jpg " diff --git a/examples/vision/matting/modnet/python/infer.py b/examples/vision/matting/modnet/python/infer.py index abb9be037..37c749010 100644 --- a/examples/vision/matting/modnet/python/infer.py +++ b/examples/vision/matting/modnet/python/infer.py @@ -56,7 +56,7 @@ result = model.predict(im) print(result) # 可视化结果 vis_im = fd.vision.vis_matting_alpha(im, result) -vis_im_with_bg = fd.vision.swap_background_matting(im, bg, result) +vis_im_with_bg = fd.vision.swap_background(im, bg, result) cv2.imwrite("visualized_result_fg.jpg", vis_im) cv2.imwrite("visualized_result_replaced_bg.jpg", vis_im_with_bg) print( diff --git a/examples/vision/matting/ppmatting/cpp/infer.cc b/examples/vision/matting/ppmatting/cpp/infer.cc index 9c917a77f..2acb2a8ca 100644 --- a/examples/vision/matting/ppmatting/cpp/infer.cc +++ b/examples/vision/matting/ppmatting/cpp/infer.cc @@ -35,16 +35,15 @@ void CpuInfer(const std::string& model_dir, const std::string& image_file, } auto im = cv::imread(image_file); - auto im_bak = im.clone(); cv::Mat bg = cv::imread(background_file); fastdeploy::vision::MattingResult res; if (!model.Predict(&im, &res)) { std::cerr << "Failed to predict." << std::endl; return; } - auto vis_im = fastdeploy::vision::VisMatting(im_bak, res); + auto vis_im = fastdeploy::vision::VisMatting(im, res); auto vis_im_with_bg = - fastdeploy::vision::SwapBackgroundMatting(im_bak, bg, res); + fastdeploy::vision::SwapBackground(im, bg, res); cv::imwrite("visualized_result.jpg", vis_im_with_bg); cv::imwrite("visualized_result_fg.jpg", vis_im); std::cout << "Visualized result save in ./visualized_result_replaced_bg.jpg " @@ -69,16 +68,15 @@ void GpuInfer(const std::string& model_dir, const std::string& image_file, } auto im = cv::imread(image_file); - auto im_bak = im.clone(); cv::Mat bg = cv::imread(background_file); fastdeploy::vision::MattingResult res; if (!model.Predict(&im, &res)) { std::cerr << "Failed to predict." << std::endl; return; } - auto vis_im = fastdeploy::vision::VisMatting(im_bak, res); + auto vis_im = fastdeploy::vision::VisMatting(im, res); auto vis_im_with_bg = - fastdeploy::vision::SwapBackgroundMatting(im_bak, bg, res); + fastdeploy::vision::SwapBackground(im, bg, res); cv::imwrite("visualized_result.jpg", vis_im_with_bg); cv::imwrite("visualized_result_fg.jpg", vis_im); std::cout << "Visualized result save in ./visualized_result_replaced_bg.jpg " @@ -104,16 +102,15 @@ void TrtInfer(const std::string& model_dir, const std::string& image_file, } auto im = cv::imread(image_file); - auto im_bak = im.clone(); cv::Mat bg = cv::imread(background_file); fastdeploy::vision::MattingResult res; if (!model.Predict(&im, &res)) { std::cerr << "Failed to predict." << std::endl; return; } - auto vis_im = fastdeploy::vision::VisMatting(im_bak, res); + auto vis_im = fastdeploy::vision::VisMatting(im, res); auto vis_im_with_bg = - fastdeploy::vision::SwapBackgroundMatting(im_bak, bg, res); + fastdeploy::vision::SwapBackground(im, bg, res); cv::imwrite("visualized_result.jpg", vis_im_with_bg); cv::imwrite("visualized_result_fg.jpg", vis_im); std::cout << "Visualized result save in ./visualized_result_replaced_bg.jpg " diff --git a/examples/vision/matting/ppmatting/python/infer.py b/examples/vision/matting/ppmatting/python/infer.py index 32a81eca7..89913cd11 100644 --- a/examples/vision/matting/ppmatting/python/infer.py +++ b/examples/vision/matting/ppmatting/python/infer.py @@ -60,7 +60,7 @@ result = model.predict(im) print(result) # 可视化结果 vis_im = fd.vision.vis_matting(im, result) -vis_im_with_bg = fd.vision.swap_background_matting(im, bg, result) +vis_im_with_bg = fd.vision.swap_background(im, bg, result) cv2.imwrite("visualized_result_fg.jpg", vis_im) cv2.imwrite("visualized_result_replaced_bg.jpg", vis_im_with_bg) print( diff --git a/examples/vision/matting/rvm/python/infer.py b/examples/vision/matting/rvm/python/infer.py index fcde64fb2..0e9eb6b21 100755 --- a/examples/vision/matting/rvm/python/infer.py +++ b/examples/vision/matting/rvm/python/infer.py @@ -86,7 +86,7 @@ if args.video is not None: break result = model.predict(frame) vis_im = fd.vision.vis_matting(frame, result) - vis_im_with_bg = fd.vision.swap_background_matting(frame, bg, result) + vis_im_with_bg = fd.vision.swap_background(frame, bg, result) alpha.write(vis_im) composition.write(vis_im_with_bg) cv2.waitKey(30) @@ -104,7 +104,7 @@ if args.image is not None: print(result) # 可视化结果 vis_im = fd.vision.vis_matting(im, result) - vis_im_with_bg = fd.vision.swap_background_matting(im, bg, result) + vis_im_with_bg = fd.vision.swap_background(im, bg, result) cv2.imwrite("visualized_result_fg.jpg", vis_im) cv2.imwrite("visualized_result_replaced_bg.jpg", vis_im_with_bg) print( From 37590ad2dc00abed74f043ce848a3dd179cb451a Mon Sep 17 00:00:00 2001 From: huangjianhui <852142024@qq.com> Date: Mon, 12 Dec 2022 21:29:57 +0800 Subject: [PATCH 05/30] Update README_CN.md --- README_CN.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README_CN.md b/README_CN.md index 8814215fe..d296e37b3 100755 --- a/README_CN.md +++ b/README_CN.md @@ -118,7 +118,7 @@ model = vision.detection.PPYOLOE("ppyoloe_crn_l_300e_coco/model.pdmodel", "ppyoloe_crn_l_300e_coco/model.pdiparams", "ppyoloe_crn_l_300e_coco/infer_cfg.yml") im = cv2.imread("000000014439.jpg") -result = model.predict(im.copy()) +result = model.predict(im) print(result) vis_im = vision.vis_detection(im, result, score_threshold=0.5) From 8d038e430856110fa666736fd456a6fda0c40464 Mon Sep 17 00:00:00 2001 From: huangjianhui <852142024@qq.com> Date: Mon, 12 Dec 2022 21:36:15 +0800 Subject: [PATCH 06/30] Update README_CN.md --- README_CN.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/README_CN.md b/README_CN.md index d296e37b3..30dc326f8 100755 --- a/README_CN.md +++ b/README_CN.md @@ -165,7 +165,7 @@ int main(int argc, char* argv[]) { vision::DetectionResult res; model.Predict(&im, &res); - auto vis_im = vision::Visualize::VisDetection(im, res, 0.5); + auto vis_im = vision::VisDetection(im, res, 0.5); cv::imwrite("vis_image.jpg", vis_im); return 0; } From b016b762cf3f9bcd6451c54cdfe16f569feed867 Mon Sep 17 00:00:00 2001 From: huangjianhui <852142024@qq.com> Date: Wed, 18 Jan 2023 16:40:07 +0800 Subject: [PATCH 07/30] Update preprocessor.h --- fastdeploy/vision/segmentation/ppseg/preprocessor.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/fastdeploy/vision/segmentation/ppseg/preprocessor.h b/fastdeploy/vision/segmentation/ppseg/preprocessor.h index 6452e8e0e..1b27863e4 100644 --- a/fastdeploy/vision/segmentation/ppseg/preprocessor.h +++ b/fastdeploy/vision/segmentation/ppseg/preprocessor.h @@ -31,7 +31,8 @@ class FASTDEPLOY_DECL PaddleSegPreprocessor { /** \brief Process the input image and prepare input tensors for runtime * * \param[in] images The input image data list, all the elements are returned by cv::imread() - * \param[in] outputs The output tensors which will feed in runtime, include image + * \param[in] outputs The output tensors which will feed in runtime + * \param[in] imgs_info The original input images shape info map, key is "shape_info", value is vector> a{{height, width}} * \return true if the preprocess successed, otherwise false */ virtual bool Run( From 03cdc4cde36056f48bcd1b9bbd41e4ba3e6b07fa Mon Sep 17 00:00:00 2001 From: felixhjh <852142024@qq.com> Date: Wed, 18 Jan 2023 12:20:05 +0000 Subject: [PATCH 08/30] PaddleSeg supports triton serving --- .../serving/models/paddleseg/1/README.md | 3 + .../serving/models/paddleseg/config.pbtxt | 67 ++++++++++ .../serving/models/postprocess/1/model.py | 115 +++++++++++++++++ .../serving/models/postprocess/config.pbtxt | 30 +++++ .../serving/models/preprocess/1/deploy.yaml | 12 ++ .../serving/models/preprocess/1/model.py | 117 ++++++++++++++++++ .../serving/models/preprocess/config.pbtxt | 34 +++++ .../serving/models/runtime/1/README.md | 5 + .../serving/models/runtime/config.pbtxt | 60 +++++++++ 9 files changed, 443 insertions(+) create mode 100644 examples/vision/segmentation/paddleseg/serving/models/paddleseg/1/README.md create mode 100644 examples/vision/segmentation/paddleseg/serving/models/paddleseg/config.pbtxt create mode 100755 examples/vision/segmentation/paddleseg/serving/models/postprocess/1/model.py create mode 100644 examples/vision/segmentation/paddleseg/serving/models/postprocess/config.pbtxt create mode 100644 examples/vision/segmentation/paddleseg/serving/models/preprocess/1/deploy.yaml create mode 100644 examples/vision/segmentation/paddleseg/serving/models/preprocess/1/model.py create mode 100644 examples/vision/segmentation/paddleseg/serving/models/preprocess/config.pbtxt create mode 100644 examples/vision/segmentation/paddleseg/serving/models/runtime/1/README.md create mode 100644 examples/vision/segmentation/paddleseg/serving/models/runtime/config.pbtxt diff --git a/examples/vision/segmentation/paddleseg/serving/models/paddleseg/1/README.md b/examples/vision/segmentation/paddleseg/serving/models/paddleseg/1/README.md new file mode 100644 index 000000000..42ae7e483 --- /dev/null +++ b/examples/vision/segmentation/paddleseg/serving/models/paddleseg/1/README.md @@ -0,0 +1,3 @@ +# PaddleSeg Pipeline + +The pipeline directory does not have model files, but a version number directory needs to be maintained. diff --git a/examples/vision/segmentation/paddleseg/serving/models/paddleseg/config.pbtxt b/examples/vision/segmentation/paddleseg/serving/models/paddleseg/config.pbtxt new file mode 100644 index 000000000..9571a5b91 --- /dev/null +++ b/examples/vision/segmentation/paddleseg/serving/models/paddleseg/config.pbtxt @@ -0,0 +1,67 @@ +platform: "ensemble" + +input [ + { + name: "INPUT" + data_type: TYPE_UINT8 + dims: [-1, -1, -1, 3 ] + } +] + +output [ + { + name: "SEG_RESULT" + data_type: TYPE_STRING + dims: [ -1 ] + } +] + +ensemble_scheduling { + step [ + { + model_name: "preprocess" + model_version: 1 + input_map { + key: "preprocess_input" + value: "INPUT" + } + output_map { + key: "preprocess_output_1" + value: "RUNTIME_INPUT_1" + } + output_map { + key: "preprocess_output_2" + value: "POSTPROCESS_INPUT_2" + } + }, + { + model_name: "runtime" + model_version: 1 + input_map { + key: "x" + value: "RUNTIME_INPUT_1" + } + output_map { + key: "argmax_0.tmp_0" + value: "RUNTIME_OUTPUT" + } + }, + { + model_name: "postprocess" + model_version: 1 + input_map { + key: "post_input_1" + value: "RUNTIME_OUTPUT" + } + input_map { + key: "post_input_2" + value: "POSTPROCESS_INPUT_2" + } + output_map { + key: "post_output" + value: "SEG_RESULT" + } + } + ] +} + diff --git a/examples/vision/segmentation/paddleseg/serving/models/postprocess/1/model.py b/examples/vision/segmentation/paddleseg/serving/models/postprocess/1/model.py new file mode 100755 index 000000000..510aad6ea --- /dev/null +++ b/examples/vision/segmentation/paddleseg/serving/models/postprocess/1/model.py @@ -0,0 +1,115 @@ +# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import json +import numpy as np +import time +import os +import fastdeploy as fd + +# triton_python_backend_utils is available in every Triton Python model. You +# need to use this module to create inference requests and responses. It also +# contains some utility functions for extracting information from model_config +# and converting Triton input/output types to numpy types. +import triton_python_backend_utils as pb_utils + + +class TritonPythonModel: + """Your Python model must use the same class name. Every Python model + that is created must have "TritonPythonModel" as the class name. + """ + + def initialize(self, args): + """`initialize` is called only once when the model is being loaded. + Implementing `initialize` function is optional. This function allows + the model to intialize any state associated with this model. + Parameters + ---------- + args : dict + Both keys and values are strings. The dictionary keys and values are: + * model_config: A JSON string containing the model configuration + * model_instance_kind: A string containing model instance kind + * model_instance_device_id: A string containing model instance device ID + * model_repository: Model repository path + * model_version: Model version + * model_name: Model name + """ + # You must parse model_config. JSON string is not parsed here + self.model_config = json.loads(args['model_config']) + print("model_config:", self.model_config) + + self.input_names = [] + for input_config in self.model_config["input"]: + self.input_names.append(input_config["name"]) + print("postprocess input names:", self.input_names) + + self.output_names = [] + self.output_dtype = [] + for output_config in self.model_config["output"]: + self.output_names.append(output_config["name"]) + dtype = pb_utils.triton_string_to_numpy(output_config["data_type"]) + self.output_dtype.append(dtype) + print("postprocess output names:", self.output_names) + + yaml_path = os.path.abspath(os.path.dirname(__file__)) + "/deploy.yaml" + self.postprocess_ = fd.vision.segmentation.PaddleSegPostprocessor( + yaml_path) + + def execute(self, requests): + """`execute` must be implemented in every Python model. `execute` + function receives a list of pb_utils.InferenceRequest as the only + argument. This function is called when an inference is requested + for this model. Depending on the batching configuration (e.g. Dynamic + Batching) used, `requests` may contain multiple requests. Every + Python model, must create one pb_utils.InferenceResponse for every + pb_utils.InferenceRequest in `requests`. If there is an error, you can + set the error argument when creating a pb_utils.InferenceResponse. + Parameters + ---------- + requests : list + A list of pb_utils.InferenceRequest + Returns + ------- + list + A list of pb_utils.InferenceResponse. The length of this list must + be the same as `requests` + """ + responses = [] + for request in requests: + infer_outputs = pb_utils.get_input_tensor_by_name( + request, self.input_names[0]) + im_info = pb_utils.get_input_tensor_by_name(request, + self.input_names[1]) + infer_outputs = infer_outputs.as_numpy() + im_info = im_info.as_numpy() + for i in range(im_info.shape[0]): + im_info[i] = json.loads(im_info[i].decode('utf-8').replace( + "'", '"')) + + results = self.postprocess_.run([infer_outputs], im_info[0]) + r_str = fd.vision.utils.fd_result_to_json(results) + + r_np = np.array(r_str, dtype=np.object_) + out_tensor = pb_utils.Tensor(self.output_names[0], r_np) + inference_response = pb_utils.InferenceResponse( + output_tensors=[out_tensor, ]) + responses.append(inference_response) + return responses + + def finalize(self): + """`finalize` is called only once when the model is being unloaded. + Implementing `finalize` function is optional. This function allows + the model to perform any necessary clean ups before exit. + """ + print('Cleaning up...') diff --git a/examples/vision/segmentation/paddleseg/serving/models/postprocess/config.pbtxt b/examples/vision/segmentation/paddleseg/serving/models/postprocess/config.pbtxt new file mode 100644 index 000000000..81f31ba08 --- /dev/null +++ b/examples/vision/segmentation/paddleseg/serving/models/postprocess/config.pbtxt @@ -0,0 +1,30 @@ +name: "postprocess" +backend: "python" + +input [ + { + name: "post_input_1" + data_type: TYPE_INT32 + dims: [-1, -1, -1] + }, + { + name: "post_input_2" + data_type: TYPE_STRING + dims: [ -1 ] + } +] + +output [ + { + name: "post_output" + data_type: TYPE_STRING + dims: [ -1 ] + } +] + +instance_group [ + { + count: 1 + kind: KIND_CPU + } +] diff --git a/examples/vision/segmentation/paddleseg/serving/models/preprocess/1/deploy.yaml b/examples/vision/segmentation/paddleseg/serving/models/preprocess/1/deploy.yaml new file mode 100644 index 000000000..6d33e5009 --- /dev/null +++ b/examples/vision/segmentation/paddleseg/serving/models/preprocess/1/deploy.yaml @@ -0,0 +1,12 @@ +Deploy: + input_shape: + - -1 + - 3 + - -1 + - -1 + model: model.pdmodel + output_dtype: int32 + output_op: argmax + params: model.pdiparams + transforms: + - type: Normalize diff --git a/examples/vision/segmentation/paddleseg/serving/models/preprocess/1/model.py b/examples/vision/segmentation/paddleseg/serving/models/preprocess/1/model.py new file mode 100644 index 000000000..48a72d6fa --- /dev/null +++ b/examples/vision/segmentation/paddleseg/serving/models/preprocess/1/model.py @@ -0,0 +1,117 @@ +# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import json +import numpy as np +import os + +import fastdeploy as fd + +# triton_python_backend_utils is available in every Triton Python model. You +# need to use this module to create inference requests and responses. It also +# contains some utility functions for extracting information from model_config +# and converting Triton input/output types to numpy types. +import triton_python_backend_utils as pb_utils + + +class TritonPythonModel: + """Your Python model must use the same class name. Every Python model + that is created must have "TritonPythonModel" as the class name. + """ + + def initialize(self, args): + """`initialize` is called only once when the model is being loaded. + Implementing `initialize` function is optional. This function allows + the model to intialize any state associated with this model. + Parameters + ---------- + args : dict + Both keys and values are strings. The dictionary keys and values are: + * model_config: A JSON string containing the model configuration + * model_instance_kind: A string containing model instance kind + * model_instance_device_id: A string containing model instance device ID + * model_repository: Model repository path + * model_version: Model version + * model_name: Model name + """ + # You must parse model_config. JSON string is not parsed here + self.model_config = json.loads(args['model_config']) + print("model_config:", self.model_config) + + self.input_names = [] + for input_config in self.model_config["input"]: + self.input_names.append(input_config["name"]) + print("preprocess input names:", self.input_names) + + self.output_names = [] + self.output_dtype = [] + for output_config in self.model_config["output"]: + self.output_names.append(output_config["name"]) + # dtype = pb_utils.triton_string_to_numpy(output_config["data_type"]) + # self.output_dtype.append(dtype) + self.output_dtype.append(output_config["data_type"]) + print("preprocess output names:", self.output_names) + + # init PaddleSegPreprocess class + yaml_path = os.path.abspath(os.path.dirname(__file__)) + "/deploy.yaml" + self.preprocess_ = fd.vision.segmentation.PaddleSegPreprocessor( + yaml_path) + #if args['model_instance_kind'] == 'GPU': + # device_id = int(args['model_instance_device_id']) + # self.preprocess_.use_gpu(device_id) + + def execute(self, requests): + """`execute` must be implemented in every Python model. `execute` + function receives a list of pb_utils.InferenceRequest as the only + argument. This function is called when an inference is requested + for this model. Depending on the batching configuration (e.g. Dynamic + Batching) used, `requests` may contain multiple requests. Every + Python model, must create one pb_utils.InferenceResponse for every + pb_utils.InferenceRequest in `requests`. If there is an error, you can + set the error argument when creating a pb_utils.InferenceResponse. + Parameters + ---------- + requests : list + A list of pb_utils.InferenceRequest + Returns + ------- + list + A list of pb_utils.InferenceResponse. The length of this list must + be the same as `requests` + """ + responses = [] + for request in requests: + data = pb_utils.get_input_tensor_by_name(request, + self.input_names[0]) + data = data.as_numpy() + outputs, im_info = self.preprocess_.run(data) + + # PaddleSeg preprocess has two outputs + dlpack_tensor = outputs[0].to_dlpack() + output_tensor_0 = pb_utils.Tensor.from_dlpack(self.output_names[0], + dlpack_tensor) + output_tensor_1 = pb_utils.Tensor( + self.output_names[1], np.array( + [im_info], dtype=np.object_)) + inference_response = pb_utils.InferenceResponse( + output_tensors=[output_tensor_0, output_tensor_1]) + responses.append(inference_response) + return responses + + def finalize(self): + """`finalize` is called only once when the model is being unloaded. + Implementing `finalize` function is optional. This function allows + the model to perform any necessary clean ups before exit. + """ + print('Cleaning up...') diff --git a/examples/vision/segmentation/paddleseg/serving/models/preprocess/config.pbtxt b/examples/vision/segmentation/paddleseg/serving/models/preprocess/config.pbtxt new file mode 100644 index 000000000..01cb94869 --- /dev/null +++ b/examples/vision/segmentation/paddleseg/serving/models/preprocess/config.pbtxt @@ -0,0 +1,34 @@ +name: "preprocess" +backend: "python" + +input [ + { + name: "preprocess_input" + data_type: TYPE_UINT8 + dims: [-1, -1, -1, 3 ] + } +] + +output [ + { + name: "preprocess_output_1" + data_type: TYPE_FP32 + dims: [-1, 3, -1, -1 ] + }, + { + name: "preprocess_output_2" + data_type: TYPE_STRING + dims: [ -1] + } +] + +instance_group [ + { + # The number of instances is 1 + count: 1 + # Use CPU, GPU inference option is:KIND_GPU + kind: KIND_CPU + # The instance is deployed on the 0th GPU card + # gpus: [0] + } +] diff --git a/examples/vision/segmentation/paddleseg/serving/models/runtime/1/README.md b/examples/vision/segmentation/paddleseg/serving/models/runtime/1/README.md new file mode 100644 index 000000000..1e5d914b4 --- /dev/null +++ b/examples/vision/segmentation/paddleseg/serving/models/runtime/1/README.md @@ -0,0 +1,5 @@ +# Runtime Directory + +This directory holds the model files. +Paddle models must be model.pdmodel and model.pdiparams files. +ONNX models must be model.onnx files. diff --git a/examples/vision/segmentation/paddleseg/serving/models/runtime/config.pbtxt b/examples/vision/segmentation/paddleseg/serving/models/runtime/config.pbtxt new file mode 100644 index 000000000..bd145c590 --- /dev/null +++ b/examples/vision/segmentation/paddleseg/serving/models/runtime/config.pbtxt @@ -0,0 +1,60 @@ +# optional, If name is specified it must match the name of the model repository directory containing the model. +name: "runtime" +backend: "fastdeploy" + +# Input configuration of the model +input [ + { + # input name + name: "x" + # input type such as TYPE_FP32、TYPE_UINT8、TYPE_INT8、TYPE_INT16、TYPE_INT32、TYPE_INT64、TYPE_FP16、TYPE_STRING + data_type: TYPE_FP32 + # input shape + dims: [-1, 3, -1, -1 ] + } +] + +# The output of the model is configured in the same format as the input +output [ + { + name: "argmax_0.tmp_0" + data_type: TYPE_INT32 + dims: [ -1, -1, -1 ] + } +] + +# Number of instances of the model +instance_group [ + { + # The number of instances is 1 + count: 1 + # Use GPU, CPU inference option is:KIND_CPU + kind: KIND_GPU + # The instance is deployed on the 0th GPU card + gpus: [0] + } +] + +optimization { + execution_accelerators { + gpu_execution_accelerator : [ { + # use TRT engine + name: "paddle", + #name: "tensorrt", + # use fp16 on TRT engine + parameters { key: "precision" value: "trt_fp32" } + }, + { + name: "min_shape" + parameters { key: "x" value: "1 3 256 256" } + }, + { + name: "opt_shape" + parameters { key: "x" value: "1 3 1024 1024" } + }, + { + name: "max_shape" + parameters { key: "x" value: "16 3 2048 2048" } + } + ] +}} From 1a3d0f86217feebd6d006b6091e44638ad7df9c0 Mon Sep 17 00:00:00 2001 From: felixhjh <852142024@qq.com> Date: Wed, 18 Jan 2023 13:23:50 +0000 Subject: [PATCH 09/30] Add PaddleSeg simple serving example --- .../paddleseg/python/serving/README.md | 36 ++++++++++++++++++ .../paddleseg/python/serving/README_CN.md | 36 ++++++++++++++++++ .../paddleseg/python/serving/client.py | 23 +++++++++++ .../paddleseg/python/serving/server.py | 38 +++++++++++++++++++ 4 files changed, 133 insertions(+) create mode 100644 examples/vision/segmentation/paddleseg/python/serving/README.md create mode 100644 examples/vision/segmentation/paddleseg/python/serving/README_CN.md create mode 100644 examples/vision/segmentation/paddleseg/python/serving/client.py create mode 100644 examples/vision/segmentation/paddleseg/python/serving/server.py diff --git a/examples/vision/segmentation/paddleseg/python/serving/README.md b/examples/vision/segmentation/paddleseg/python/serving/README.md new file mode 100644 index 000000000..da41a3a00 --- /dev/null +++ b/examples/vision/segmentation/paddleseg/python/serving/README.md @@ -0,0 +1,36 @@ +English | [简体中文](README_CN.md) + +# PaddleSegmentation Python Simple Serving Demo + + +## Environment + +- 1. Prepare environment and install FastDeploy Python whl, refer to [download_prebuilt_libraries](../../../../../../docs/en/build_and_install/download_prebuilt_libraries.md) + +Server: +```bash +# Download demo code +git clone https://github.com/PaddlePaddle/FastDeploy.git +cd FastDeploy/examples/vision/segmentation/paddleseg/python/serving + +# Download PP_LiteSeg model +wget https://bj.bcebos.com/paddlehub/fastdeploy/PP_LiteSeg_B_STDC2_cityscapes_with_argmax_infer.tgz +tar -xvf PP_LiteSeg_B_STDC2_cityscapes_with_argmax_infer.tgz + +# Launch server, change the configurations in server.py to select hardware, backend, etc. +# and use --host, --port to specify IP and port +fastdeploy simple_serving --app server:app +``` + +Client: +```bash +# Download demo code +git clone https://github.com/PaddlePaddle/FastDeploy.git +cd FastDeploy/examples/vision/segmentation/paddleseg/python/serving + +# Download test image +wget https://paddleseg.bj.bcebos.com/dygraph/demo/cityscapes_demo.png + +# Send request and get inference result (Please adapt the IP and port if necessary) +python client.py +``` diff --git a/examples/vision/segmentation/paddleseg/python/serving/README_CN.md b/examples/vision/segmentation/paddleseg/python/serving/README_CN.md new file mode 100644 index 000000000..3f382c904 --- /dev/null +++ b/examples/vision/segmentation/paddleseg/python/serving/README_CN.md @@ -0,0 +1,36 @@ +简体中文 | [English](README.md) + +# PaddleSegmentation Python轻量服务化部署示例 + +在部署前，需确认以下两个步骤 + +- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../../docs/cn/build_and_install/download_prebuilt_libraries.md) +- 2. FastDeploy Python whl包安装，参考[FastDeploy Python安装](../../../../../../docs/cn/build_and_install/download_prebuilt_libraries.md) + +服务端： +```bash +# 下载部署示例代码 +git clone https://github.com/PaddlePaddle/FastDeploy.git +cd FastDeploy/examples/vision/segmentation/paddleseg/python/serving + +# 下载PP_LiteSeg模型文件 +wget https://bj.bcebos.com/paddlehub/fastdeploy/PP_LiteSeg_B_STDC2_cityscapes_with_argmax_infer.tgz +tar -xvf PP_LiteSeg_B_STDC2_cityscapes_with_argmax_infer.tgz + +# 启动服务，可修改server.py中的配置项来指定硬件、后端等 +# 可通过--host、--port指定IP和端口号 +fastdeploy simple_serving --app server:app +``` + +客户端： +```bash +# 下载部署示例代码 +git clone https://github.com/PaddlePaddle/FastDeploy.git +cd FastDeploy/examples/vision/detection/paddledetection/python/serving + +# 下载测试图片 +wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg + +# 请求服务，获取推理结果（如有必要，请修改脚本中的IP和端口号） +python client.py +``` diff --git a/examples/vision/segmentation/paddleseg/python/serving/client.py b/examples/vision/segmentation/paddleseg/python/serving/client.py new file mode 100644 index 000000000..e652c4462 --- /dev/null +++ b/examples/vision/segmentation/paddleseg/python/serving/client.py @@ -0,0 +1,23 @@ +import requests +import json +import cv2 +import fastdeploy as fd +from fastdeploy.serving.utils import cv2_to_base64 + +if __name__ == '__main__': + url = "http://127.0.0.1:8000/fd/ppliteseg" + headers = {"Content-Type": "application/json"} + + im = cv2.imread("cityscapes_demo.png") + data = {"data": {"image": cv2_to_base64(im)}, "parameters": {}} + + resp = requests.post(url=url, headers=headers, data=json.dumps(data)) + if resp.status_code == 200: + r_json = json.loads(resp.json()["result"]) + result = fd.vision.utils.json_to_segmentation(r_json) + vis_im = fd.vision.vis_segmentation(im, result, weight=0.5) + cv2.imwrite("visualized_result.jpg", vis_im) + print("Visualized result save in ./visualized_result.jpg") + else: + print("Error code:", resp.status_code) + print(resp.text) diff --git a/examples/vision/segmentation/paddleseg/python/serving/server.py b/examples/vision/segmentation/paddleseg/python/serving/server.py new file mode 100644 index 000000000..2ae2df09c --- /dev/null +++ b/examples/vision/segmentation/paddleseg/python/serving/server.py @@ -0,0 +1,38 @@ +import fastdeploy as fd +from fastdeploy.serving.server import SimpleServer +import os +import logging + +logging.getLogger().setLevel(logging.INFO) + +# Configurations +model_dir = 'PP_LiteSeg_B_STDC2_cityscapes_with_argmax_infer' +device = 'cpu' +use_trt = False + +# Prepare model +model_file = os.path.join(model_dir, "model.pdmodel") +params_file = os.path.join(model_dir, "model.pdiparams") +config_file = os.path.join(model_dir, "deploy.yaml") + +# Setup runtime option to select hardware, backend, etc. +option = fd.RuntimeOption() +if device.lower() == 'gpu': + option.use_gpu() +if use_trt: + option.use_trt_backend() + option.set_trt_cache_file('pp_lite_seg.trt') + +# Create model instance +model_instance = fd.vision.segmentation.PaddleSegModel( + model_file=model_file, + params_file=params_file, + config_file=config_file, + runtime_option=option) + +# Create server, setup REST API +app = SimpleServer() +app.register( + task_name="fd/ppliteseg", + model_handler=fd.serving.handler.VisionModelHandler, + predictor=model_instance) From 03fd1aa5cff4cb8adaae1c8a29aed404a38f2283 Mon Sep 17 00:00:00 2001 From: felixhjh <852142024@qq.com> Date: Wed, 18 Jan 2023 13:25:00 +0000 Subject: [PATCH 10/30] Add PaddleSeg triton serving client code --- .../serving/paddleseg_grpc_client.py | 116 ++++++++++++++++++ 1 file changed, 116 insertions(+) create mode 100644 examples/vision/segmentation/paddleseg/serving/paddleseg_grpc_client.py diff --git a/examples/vision/segmentation/paddleseg/serving/paddleseg_grpc_client.py b/examples/vision/segmentation/paddleseg/serving/paddleseg_grpc_client.py new file mode 100644 index 000000000..1fe3828db --- /dev/null +++ b/examples/vision/segmentation/paddleseg/serving/paddleseg_grpc_client.py @@ -0,0 +1,116 @@ +import logging +import numpy as np +import time +from typing import Optional +import cv2 +import json + +from tritonclient import utils as client_utils +from tritonclient.grpc import InferenceServerClient, InferInput, InferRequestedOutput, service_pb2_grpc, service_pb2 + +LOGGER = logging.getLogger("run_inference_on_triton") + + +class SyncGRPCTritonRunner: + DEFAULT_MAX_RESP_WAIT_S = 120 + + def __init__( + self, + server_url: str, + model_name: str, + model_version: str, + *, + verbose=False, + resp_wait_s: Optional[float]=None, ): + self._server_url = server_url + self._model_name = model_name + self._model_version = model_version + self._verbose = verbose + self._response_wait_t = self.DEFAULT_MAX_RESP_WAIT_S if resp_wait_s is None else resp_wait_s + + self._client = InferenceServerClient( + self._server_url, verbose=self._verbose) + error = self._verify_triton_state(self._client) + if error: + raise RuntimeError( + f"Could not communicate to Triton Server: {error}") + + LOGGER.debug( + f"Triton server {self._server_url} and model {self._model_name}:{self._model_version} " + f"are up and ready!") + + model_config = self._client.get_model_config(self._model_name, + self._model_version) + model_metadata = self._client.get_model_metadata(self._model_name, + self._model_version) + LOGGER.info(f"Model config {model_config}") + LOGGER.info(f"Model metadata {model_metadata}") + + for tm in model_metadata.inputs: + print("tm:", tm) + self._inputs = {tm.name: tm for tm in model_metadata.inputs} + self._input_names = list(self._inputs) + self._outputs = {tm.name: tm for tm in model_metadata.outputs} + self._output_names = list(self._outputs) + self._outputs_req = [ + InferRequestedOutput(name) for name in self._outputs + ] + + def Run(self, inputs): + """ + Args: + inputs: list, Each value corresponds to an input name of self._input_names + Returns: + results: dict, {name : numpy.array} + """ + infer_inputs = [] + for idx, data in enumerate(inputs): + infer_input = InferInput(self._input_names[idx], data.shape, + "UINT8") + infer_input.set_data_from_numpy(data) + infer_inputs.append(infer_input) + + results = self._client.infer( + model_name=self._model_name, + model_version=self._model_version, + inputs=infer_inputs, + outputs=self._outputs_req, + client_timeout=self._response_wait_t, ) + results = {name: results.as_numpy(name) for name in self._output_names} + return results + + def _verify_triton_state(self, triton_client): + if not triton_client.is_server_live(): + return f"Triton server {self._server_url} is not live" + elif not triton_client.is_server_ready(): + return f"Triton server {self._server_url} is not ready" + elif not triton_client.is_model_ready(self._model_name, + self._model_version): + return f"Model {self._model_name}:{self._model_version} is not ready" + return None + + +if __name__ == "__main__": + model_name = "paddleseg" + model_version = "1" + url = "localhost:8001" + runner = SyncGRPCTritonRunner(url, model_name, model_version) + im = cv2.imread("cityscapes_demo.png") + im = np.array([im, ]) + # batch input + # im = np.array([im, im, im]) + for i in range(1): + result = runner.Run([im, ]) + for name, values in result.items(): + print("output_name:", name) + # values is batch + for value in values: + value = json.loads(value) + #print(value) + import fastdeploy as fd + result = fd.C.vision.SegmentationResult() + result.label_map = value["label_map"] + result.shape = value["shape"] + im = cv2.imread("cityscapes_demo.png") + vis_im = fd.vision.vis_segmentation(im, result, 0.5) + cv2.imwrite("seg.png", vis_im) From 9d30be35a49628c36bf601c0e09ca31edbac5d69 Mon Sep 17 00:00:00 2001 From: felixhjh <852142024@qq.com> Date: Thu, 19 Jan 2023 03:37:14 +0000 Subject: [PATCH 11/30] Update triton serving runtime config.pbtxt --- .../paddleseg/serving/models/runtime/config.pbtxt | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) diff --git a/examples/vision/segmentation/paddleseg/serving/models/runtime/config.pbtxt b/examples/vision/segmentation/paddleseg/serving/models/runtime/config.pbtxt index bd145c590..875086e2b 100644 --- a/examples/vision/segmentation/paddleseg/serving/models/runtime/config.pbtxt +++ b/examples/vision/segmentation/paddleseg/serving/models/runtime/config.pbtxt @@ -1,6 +1,7 @@ # optional, If name is specified it must match the name of the model repository directory containing the model. name: "runtime" backend: "fastdeploy" +max_batch_size: 1 # Input configuration of the model input [ @@ -10,7 +11,7 @@ input [ # input type such as TYPE_FP32、TYPE_UINT8、TYPE_INT8、TYPE_INT16、TYPE_INT32、TYPE_INT64、TYPE_FP16、TYPE_STRING data_type: TYPE_FP32 # input shape - dims: [-1, 3, -1, -1 ] + dims: [3, -1, -1 ] } ] @@ -19,7 +20,7 @@ output [ { name: "argmax_0.tmp_0" data_type: TYPE_INT32 - dims: [ -1, -1, -1 ] + dims: [ -1, -1 ] } ] @@ -39,9 +40,8 @@ optimization { execution_accelerators { gpu_execution_accelerator : [ { # use TRT engine - name: "paddle", - #name: "tensorrt", - # use fp16 on TRT engine + name: "tensorrt", + # use fp32 on TRT engine parameters { key: "precision" value: "trt_fp32" } }, { @@ -54,7 +54,7 @@ optimization { }, { name: "max_shape" - parameters { key: "x" value: "16 3 2048 2048" } + parameters { key: "x" value: "1 3 2048 2048" } } ] }} From 7d145b5be386b98a4c5b709b6b57d89ca8a37bd5 Mon Sep 17 00:00:00 2001 From: felixhjh <852142024@qq.com> Date: Thu, 19 Jan 2023 06:26:04 +0000 Subject: [PATCH 12/30] Update paddleseg grpc client --- .../paddleseg/serving/paddleseg_grpc_client.py | 12 ++++-------- 1 file changed, 4 insertions(+), 8 deletions(-) diff --git a/examples/vision/segmentation/paddleseg/serving/paddleseg_grpc_client.py b/examples/vision/segmentation/paddleseg/serving/paddleseg_grpc_client.py index 1fe3828db..f200dec25 100644 --- a/examples/vision/segmentation/paddleseg/serving/paddleseg_grpc_client.py +++ b/examples/vision/segmentation/paddleseg/serving/paddleseg_grpc_client.py @@ -106,11 +106,7 @@ if __name__ == "__main__": # values is batch for value in values: value = json.loads(value) - #print(value) - import fastdeploy as fd - result = fd.C.vision.SegmentationResult() - result.label_map = value["label_map"] - result.shape = value["shape"] - im = cv2.imread("cityscapes_demo.png") - vis_im = fd.vision.vis_segmentation(im, result, 0.5) - cv2.imwrite("seg.png", vis_im) + print( + "Only print the first 20 labels in label_map of SEG_RESULT") + value["label_map"] = value["label_map"][:20] + print(value) From 175c6b0f39b4eeb4e63082bff7c6ed7bce2e0624 Mon Sep 17 00:00:00 2001 From: felixhjh <852142024@qq.com> Date: Thu, 19 Jan 2023 06:27:07 +0000 Subject: [PATCH 13/30] Add paddle serving README --- .../segmentation/paddleseg/serving/README.md | 62 +++++++++++++++++ .../paddleseg/serving/README_CN.md | 68 +++++++++++++++++++ 2 files changed, 130 insertions(+) create mode 100644 examples/vision/segmentation/paddleseg/serving/README.md create mode 100644 examples/vision/segmentation/paddleseg/serving/README_CN.md diff --git a/examples/vision/segmentation/paddleseg/serving/README.md b/examples/vision/segmentation/paddleseg/serving/README.md new file mode 100644 index 000000000..a451e8730 --- /dev/null +++ b/examples/vision/segmentation/paddleseg/serving/README.md @@ -0,0 +1,62 @@ +English | [简体中文](README_CN.md) +# PaddleSegmentation Serving Deployment Demo + +## Launch Serving + +```bash +# Download demo code +git clone https://github.com/PaddlePaddle/FastDeploy.git +cd FastDeploy/examples/vision/segmentation/paddleseg/serving + +#Download PP_LiteSeg model file +wget https://bj.bcebos.com/paddlehub/fastdeploy/PP_LiteSeg_B_STDC2_cityscapes_with_argmax_infer.tgz +tar -xvf PP_LiteSeg_B_STDC2_cityscapes_with_argmax_infer.tgz + +# Move the model files to models/infer/1 +mv yolov5s.onnx models/infer/1/ + +# Pull fastdeploy image, x.y.z is FastDeploy version, example 1.0.2. +docker pull paddlepaddle/fastdeploy:x.y.z-gpu-cuda11.4-trt8.4-21.10 + +# Run the docker. The docker name is fd_serving, and the current directory is mounted as the docker's /serving directory +nvidia-docker run -it --net=host --name fd_serving -v `pwd`/:/serving paddlepaddle/fastdeploy:x.y.z-gpu-cuda11.4-trt8.4-21.10 bash + +# Start the service (Without setting the CUDA_VISIBLE_DEVICES environment variable, it will have scheduling privileges for all GPU cards) +CUDA_VISIBLE_DEVICES=0 fastdeployserver --model-repository=/serving/models --backend-config=python,shm-default-byte-size=10485760 +``` + +Output the following contents if serving is launched + +``` +...... +I0928 04:51:15.784517 206 grpc_server.cc:4117] Started GRPCInferenceService at 0.0.0.0:8001 +I0928 04:51:15.785177 206 http_server.cc:2815] Started HTTPService at 0.0.0.0:8000 +I0928 04:51:15.826578 206 http_server.cc:167] Started Metrics Service at 0.0.0.0:8002 +``` + +## Client Requests + +Execute the following command in the physical machine to send a grpc request and output the result + +``` +#Download test images +wget https://paddleseg.bj.bcebos.com/dygraph/demo/cityscapes_demo.png + +#Installing client-side dependencies +python3 -m pip install tritonclient\[all\] + +# Send requests +python3 paddleseg_grpc_client.py +``` + +When the request is sent successfully, the results are returned in json format and printed out: + +``` + +``` + +## Modify Configs + + + +The default is to run ONNXRuntime on CPU. If developers need to run it on GPU or other inference engines, please see the [Configs File](../../../../../serving/docs/EN/model_configuration-en.md) to modify the configs in `models/runtime/config.pbtxt`. diff --git a/examples/vision/segmentation/paddleseg/serving/README_CN.md b/examples/vision/segmentation/paddleseg/serving/README_CN.md new file mode 100644 index 000000000..676272f2e --- /dev/null +++ b/examples/vision/segmentation/paddleseg/serving/README_CN.md @@ -0,0 +1,68 @@ +[English](README.md) | 简体中文 +# PaddleSegmentation 服务化部署示例 + +在服务化部署前，需确认 + +- 1. 服务化镜像的软硬件环境要求和镜像拉取命令请参考[FastDeploy服务化部署](../../../../../serving/README_CN.md) + + +## 启动服务 + +```bash +#下载部署示例代码 +git clone https://github.com/PaddlePaddle/FastDeploy.git +cd FastDeploy/examples/vision/segmentation/paddleseg/serving + +#下载yolov5模型文件 +wget https://bj.bcebos.com/paddlehub/fastdeploy/PP_LiteSeg_B_STDC2_cityscapes_with_argmax_infer.tgz +tar -xvf PP_LiteSeg_B_STDC2_cityscapes_with_argmax_infer.tgz + +# 将模型文件放入 models/runtime/1目录下 +mv PP_LiteSeg_B_STDC2_cityscapes_with_argmax_infer/model.pdmodel models/runtime/1/ +mv PP_LiteSeg_B_STDC2_cityscapes_with_argmax_infer/model.pdiparams models/runtime/1/ + +# 拉取fastdeploy镜像(x.y.z为镜像版本号，需参照serving文档替换为数字) +# GPU镜像 +docker pull registry.baidubce.com/paddlepaddle/fastdeploy:x.y.z-gpu-cuda11.4-trt8.4-21.10 +# CPU镜像 +docker pull registry.baidubce.com/paddlepaddle/fastdeploy:x.y.z-cpu-only-21.10 + +# 运行容器.容器名字为 fd_serving, 并挂载当前目录为容器的 /serving 目录 +nvidia-docker run -it --net=host --name fd_serving -v `pwd`/:/serving registry.baidubce.com/paddlepaddle/fastdeploy:x.y.z-gpu-cuda11.4-trt8.4-21.10 bash + +# 启动服务(不设置CUDA_VISIBLE_DEVICES环境变量，会拥有所有GPU卡的调度权限) +CUDA_VISIBLE_DEVICES=0 fastdeployserver --model-repository=/serving/models --backend-config=python,shm-default-byte-size=10485760 +``` +>> **注意**: 当出现"Address already in use", 请使用`--grpc-port`指定端口号来启动服务，同时更改paddleseg_grpc_client.py中的请求端口号 + +服务启动成功后， 会有以下输出: +``` +...... +I0928 04:51:15.784517 206 grpc_server.cc:4117] Started GRPCInferenceService at 0.0.0.0:8001 +I0928 04:51:15.785177 206 http_server.cc:2815] Started HTTPService at 0.0.0.0:8000 +I0928 04:51:15.826578 206 http_server.cc:167] Started Metrics Service at 0.0.0.0:8002 +``` + + +## 客户端请求 + +在物理机器中执行以下命令，发送grpc请求并输出结果 +``` +#下载测试图片 +wget https://paddleseg.bj.bcebos.com/dygraph/demo/cityscapes_demo.png + +#安装客户端依赖 +python3 -m pip install tritonclient[all] + +# 发送请求 +python3 paddleseg_grpc_client.py +``` + +发送请求成功后，会返回json格式的检测结果并打印输出: +``` + +``` + +## 配置修改 + +当前默认配置在CPU上运行ONNXRuntime引擎， 如果要在GPU或其他推理引擎上运行。 需要修改`models/runtime/config.pbtxt`中配置，详情请参考[配置文档](../../../../../serving/docs/zh_CN/model_configuration.md) From 4b293a89de7de52f40ad1f98e2819488f69d051f Mon Sep 17 00:00:00 2001 From: wwbitejotunn Date: Thu, 9 Feb 2023 05:51:30 +0000 Subject: [PATCH 14/30] fix paddle backend --- fastdeploy/runtime/backends/paddle/paddle_backend.cc | 1 + 1 file changed, 1 insertion(+) diff --git a/fastdeploy/runtime/backends/paddle/paddle_backend.cc b/fastdeploy/runtime/backends/paddle/paddle_backend.cc index e0e908c36..19493f90b 100644 --- a/fastdeploy/runtime/backends/paddle/paddle_backend.cc +++ b/fastdeploy/runtime/backends/paddle/paddle_backend.cc @@ -44,6 +44,7 @@ void PaddleBackend::BuildOption(const PaddleBackendOption& option) { "file will save to the directory where paddle model saved." << std::endl; use_static = true; + config_.SetOptimCacheDir(option.trt_option.serialize_file); } config_.EnableTensorRtEngine(option.trt_option.max_workspace_size, option.trt_option.max_batch_size, 3, From 59c5fedc36d7e3dc5538fb385d1201a0aa184df6 Mon Sep 17 00:00:00 2001 From: wwbitejotunn Date: Thu, 9 Feb 2023 12:53:08 +0000 Subject: [PATCH 15/30] set stream by raw size_t --- fastdeploy/runtime/option_pybind.cc | 1 + fastdeploy/runtime/runtime_option.cc | 4 ++++ fastdeploy/runtime/runtime_option.h | 2 ++ 3 files changed, 7 insertions(+) diff --git a/fastdeploy/runtime/option_pybind.cc b/fastdeploy/runtime/option_pybind.cc index 1c786459b..d5a6187c8 100644 --- a/fastdeploy/runtime/option_pybind.cc +++ b/fastdeploy/runtime/option_pybind.cc @@ -48,6 +48,7 @@ void BindOption(pybind11::module& m) { .def_readwrite("poros_option", &RuntimeOption::poros_option) .def_readwrite("paddle_infer_option", &RuntimeOption::paddle_infer_option) .def("set_external_stream", &RuntimeOption::SetExternalStream) + .def("set_external_raw_stream", &RuntimeOption::SetExternalRawStream) .def("set_cpu_thread_num", &RuntimeOption::SetCpuThreadNum) .def("use_paddle_backend", &RuntimeOption::UsePaddleBackend) .def("use_poros_backend", &RuntimeOption::UsePorosBackend) diff --git a/fastdeploy/runtime/runtime_option.cc b/fastdeploy/runtime/runtime_option.cc index 7538f3ea6..97b7acea7 100644 --- a/fastdeploy/runtime/runtime_option.cc +++ b/fastdeploy/runtime/runtime_option.cc @@ -93,6 +93,10 @@ void RuntimeOption::SetExternalStream(void* external_stream) { external_stream_ = external_stream; } +void RuntimeOption::SetExternalRawStream(size_t external_stream) { + external_stream_ = (void*) external_stream; +} + void RuntimeOption::SetCpuThreadNum(int thread_num) { FDASSERT(thread_num > 0, "The thread_num must be greater than 0."); cpu_thread_num = thread_num; diff --git a/fastdeploy/runtime/runtime_option.h b/fastdeploy/runtime/runtime_option.h index ecb51fe2a..6f21a2860 100644 --- a/fastdeploy/runtime/runtime_option.h +++ b/fastdeploy/runtime/runtime_option.h @@ -111,6 +111,8 @@ struct FASTDEPLOY_DECL RuntimeOption { void SetExternalStream(void* external_stream); + void SetExternalRawStream(size_t external_stream); + /* * @brief Set number of cpu threads while inference on CPU, by default it will decided by the different backends */ From 942cee83d7d0c94ed0598bb040e9fe7ec3e7f873 Mon Sep 17 00:00:00 2001 From: felixhjh <852142024@qq.com> Date: Thu, 9 Feb 2023 14:04:10 +0000 Subject: [PATCH 16/30] Add decrypt function to load encrypted model --- fastdeploy/runtime/option_pybind.cc | 1 + fastdeploy/runtime/runtime.cc | 26 +++++++++++++++++++++++++- fastdeploy/runtime/runtime.h | 3 +++ fastdeploy/runtime/runtime_option.cc | 9 +++++++++ python/fastdeploy/runtime.py | 7 +++++++ 5 files changed, 45 insertions(+), 1 deletion(-) diff --git a/fastdeploy/runtime/option_pybind.cc b/fastdeploy/runtime/option_pybind.cc index 1c786459b..7af90d831 100644 --- a/fastdeploy/runtime/option_pybind.cc +++ b/fastdeploy/runtime/option_pybind.cc @@ -35,6 +35,7 @@ void BindOption(pybind11::module& m) { .def(pybind11::init()) .def("set_model_path", &RuntimeOption::SetModelPath) .def("set_model_buffer", &RuntimeOption::SetModelBuffer) + .def("set_encryption_key", &RuntimeOption::SetEncryptionKey) .def("use_gpu", &RuntimeOption::UseGpu) .def("use_cpu", &RuntimeOption::UseCpu) .def("use_rknpu2", &RuntimeOption::UseRKNPU2) diff --git a/fastdeploy/runtime/runtime.cc b/fastdeploy/runtime/runtime.cc index 70714e4f0..2a00dfda4 100644 --- a/fastdeploy/runtime/runtime.cc +++ b/fastdeploy/runtime/runtime.cc @@ -104,7 +104,31 @@ bool AutoSelectBackend(RuntimeOption& option) { bool Runtime::Init(const RuntimeOption& _option) { option = _option; - + if ("" != option.encryption_key_) { + #ifdef ENABLE_ENCRYPTION + if (option.model_from_memory_) { + option.model_file = Decrypt(option.model_file, option.encryption_key_); + if (!(option.params_file.empty())) { + option.params_file = Decrypt(option.params_file, option.encryption_key_); + } + } else { + std::string model_buffer = ""; + FDASSERT(ReadBinaryFromFile(option.model_file, &model_buffer), + "Fail to read binary from model file"); + option.model_file = Decrypt(model_buffer, option.encryption_key_); + if (!(option.params_file.empty())) { + std::string params_buffer = ""; + FDASSERT(ReadBinaryFromFile(option.params_file, ¶ms_buffer), + "Fail to read binary from parameter file"); + option.params_file = Decrypt(params_buffer, option.encryption_key_); + } + option.model_from_memory_ = true; + } + #else + FDERROR << "The FastDeploy didn't compile with encryption function." + << std::endl; + #endif + } // Choose default backend by model format and device if backend is not // specified if (option.backend == Backend::UNKNOWN) { diff --git a/fastdeploy/runtime/runtime.h b/fastdeploy/runtime/runtime.h index 6e7dc9629..772773007 100755 --- a/fastdeploy/runtime/runtime.h +++ b/fastdeploy/runtime/runtime.h @@ -23,6 +23,9 @@ #include "fastdeploy/core/fd_tensor.h" #include "fastdeploy/runtime/runtime_option.h" #include "fastdeploy/utils/perf.h" +#ifdef ENABLE_ENCRYPTION + #include "fastdeploy/encryption/include/decrypt.h" +#endif /** \brief All C++ FastDeploy APIs are defined inside this namespace * diff --git a/fastdeploy/runtime/runtime_option.cc b/fastdeploy/runtime/runtime_option.cc index c09352d58..8568b3b7f 100644 --- a/fastdeploy/runtime/runtime_option.cc +++ b/fastdeploy/runtime/runtime_option.cc @@ -36,6 +36,15 @@ void RuntimeOption::SetModelBuffer(const std::string& model_buffer, model_from_memory_ = true; } +void RuntimeOption::SetEncryptionKey(const std::string& encryption_key) { + #ifdef ENABLE_ENCRYPTION + encryption_key_ = encryption_key; + #else + FDERROR << "The FastDeploy didn't compile with encryption function." + << std::endl; + #endif +} + void RuntimeOption::UseGpu(int gpu_id) { #ifdef WITH_GPU device = Device::GPU; diff --git a/python/fastdeploy/runtime.py b/python/fastdeploy/runtime.py index cd7b6641b..1d2fc1c1d 100644 --- a/python/fastdeploy/runtime.py +++ b/python/fastdeploy/runtime.py @@ -187,6 +187,13 @@ class RuntimeOption: return self._option.set_model_buffer(model_buffer, params_buffer, model_format) + def set_encryption_key(self, + encryption_key): + """When loading encrypted model, encryption_key is required to decrypte model + :param encryption_key: (str)The key for decrypting model + """ + return self._option.set_encryption_key(encryption_key) + def use_gpu(self, device_id=0): """Inference with Nvidia GPU From 5160771a1c95035fedc608fb4ab4325f82459f05 Mon Sep 17 00:00:00 2001 From: felixhjh <852142024@qq.com> Date: Fri, 10 Feb 2023 03:24:45 +0000 Subject: [PATCH 17/30] Update runtime_option.h --- fastdeploy/runtime/runtime_option.h | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/fastdeploy/runtime/runtime_option.h b/fastdeploy/runtime/runtime_option.h index 0aa6bbec8..b263f4ad3 100644 --- a/fastdeploy/runtime/runtime_option.h +++ b/fastdeploy/runtime/runtime_option.h @@ -59,6 +59,12 @@ struct FASTDEPLOY_DECL RuntimeOption { const std::string& params_buffer = "", const ModelFormat& format = ModelFormat::PADDLE); + /** \brief When loading encrypted model, encryption_key is required to decrypte model + * + * \param[in] encryption_key The key for decrypting model + */ + void SetEncryptionKey(const std::string& encryption_key); + /// Use cpu to inference, the runtime will inference on CPU by default void UseCpu(); /// Use Nvidia GPU to inference @@ -178,6 +184,8 @@ struct FASTDEPLOY_DECL RuntimeOption { /// format of input model ModelFormat model_format = ModelFormat::PADDLE; + std::string encryption_key_ = ""; + // for cpu inference // default will let the backend choose their own default value int cpu_thread_num = -1; From 9d8b08415570d7cbb14f08c1025a52d0276d974f Mon Sep 17 00:00:00 2001 From: felixhjh <852142024@qq.com> Date: Fri, 10 Feb 2023 05:06:20 +0000 Subject: [PATCH 18/30] Update tutorials for encryption model --- tutorials/encrypt_model/README.md | 46 ++++++++++++++++++++++++++ tutorials/encrypt_model/README_CN.md | 48 ++++++++++++++++++++++++++++ tutorials/encrypt_model/encrypt.py | 33 +++++++++++++++++++ 3 files changed, 127 insertions(+) create mode 100644 tutorials/encrypt_model/README.md create mode 100644 tutorials/encrypt_model/README_CN.md create mode 100644 tutorials/encrypt_model/encrypt.py diff --git a/tutorials/encrypt_model/README.md b/tutorials/encrypt_model/README.md new file mode 100644 index 000000000..8a49c107c --- /dev/null +++ b/tutorials/encrypt_model/README.md @@ -0,0 +1,46 @@ +English | [中文](README_CN.md) + +# FastDeploy generates an encrypted model + +This directory provides `encrypt.py` to quickly complete the encryption of the model and parameter files of ResNet50_vd + +## encryption +```bash +# Download deployment example code +git clone https://github.com/PaddlePaddle/FastDeploy.git +cd FastDeploy/tutorials/encrypt_model + +# Download the ResNet50_vd model file +wget https://bj.bcebos.com/paddlehub/fastdeploy/ResNet50_vd_infer.tgz +tar -xvf ResNet50_vd_infer.tgz + +python encrypt.py --model ResNet50_vd_infer +``` +>> **Note** After the encryption is completed, the ResNet50_vd_infer_encrypt folder will be generated, including `__model__.encrypted`, `__params__.encrypted`, `encryption_key.txt` three files, where `encryption_key.txt` contains the encrypted key. At the same time, you need to copy the `inference_cls.yaml` configuration file in the original folder to the ResNet50_vd_infer_encrypt folder for subsequent deployment + +### Python encryption interface + +Use the encrypted interface through the following interface settings +```python +import fastdeploy as fd +import os +# when key is not given, key will be automatically generated. +# otherwise, the file will be encrypted by specific key +encrypted_model, key = fd.encryption.encrypt(model_file.read()) +encrypted_params, key= fd.encryption.encrypt(params_file.read(), key) +``` + +### FastDeploy deployment encryption model (decryption) + +Through the setting of the following interface, FastDeploy can deploy the encryption model +```python +import fastdeploy as fd +option = fd.RuntimeOption() +option.set_encryption_key(key) +``` + +```C++ +fastdeploy::RuntimeOption option; +option.SetEncryptionKey(key) +``` +>> **Note** For more details about RuntimeOption, please refer to [RuntimeOption Python Documentation](https://www.paddlepaddle.org.cn/fastdeploy-api-doc/python/html/runtime_option.html), [ RuntimeOption C++ Documentation](https://www.paddlepaddle.org.cn/fastdeploy-api-doc/cpp/html/structfastdeploy_1_1RuntimeOption.html) \ No newline at end of file diff --git a/tutorials/encrypt_model/README_CN.md b/tutorials/encrypt_model/README_CN.md new file mode 100644 index 000000000..8230f68e6 --- /dev/null +++ b/tutorials/encrypt_model/README_CN.md @@ -0,0 +1,48 @@ +[English](README.md) | 中文 + +# 使用FastDeploy生成加密模型 + +本目录下提供`encrypt.py`快速完成ResNet50_vd的模型和参数文件加密 + +FastDeploy支持对称加密的方案，通过调用OpenSSL中的对称加密算法（AES）对模型进行加密并产生密钥 + +## 加密 +```bash +#下载加密示例代码 +git clone https://github.com/PaddlePaddle/FastDeploy.git +cd FastDeploy/tutorials/encrypt_model + +# 下载ResNet50_vd模型文件 +wget https://bj.bcebos.com/paddlehub/fastdeploy/ResNet50_vd_infer.tgz +tar -xvf ResNet50_vd_infer.tgz + +python encrypt.py --model ResNet50_vd_infer +``` +>> **注意** 加密完成后会生成ResNet50_vd_infer_encrypt文件夹，包含`__model__.encrypted`,`__params__.encrypted`,`encryption_key.txt`三个文件，其中`encryption_key.txt`包含加密后的秘钥，同时需要将原文件夹中的、`inference_cls.yaml`配置文件 拷贝至ResNet50_vd_infer_encrypt文件夹，以便后续部署使用 + +### Python加密接口 + +通过如下接口的设定，使用加密接口（解密） +```python +import fastdeploy as fd +import os +# when key is not given, key will be automatically generated. +# otherwise, the file will be encrypted by specific key +encrypted_model, key = fd.encryption.encrypt(model_file.read()) +encrypted_params, key= fd.encryption.encrypt(params_file.read(), key) +``` + +### FastDeploy 部署加密模型 + +通过如下接口的设定，完成加密模型的推理 +```python +import fastdeploy as fd +option = fd.RuntimeOption() +option.set_encryption_key(key) +``` + +```C++ +fastdeploy::RuntimeOption option; +option.SetEncryptionKey(key) +``` +>> **注意** RuntimeOption的更多详细信息，请参考[RuntimeOption Python文档](https://www.paddlepaddle.org.cn/fastdeploy-api-doc/python/html/runtime_option.html)，[RuntimeOption C++文档](https://www.paddlepaddle.org.cn/fastdeploy-api-doc/cpp/html/structfastdeploy_1_1RuntimeOption.html) \ No newline at end of file diff --git a/tutorials/encrypt_model/encrypt.py b/tutorials/encrypt_model/encrypt.py new file mode 100644 index 000000000..380509042 --- /dev/null +++ b/tutorials/encrypt_model/encrypt.py @@ -0,0 +1,33 @@ +import fastdeploy as fd +import os + +def parse_arguments(): + import argparse + import ast + parser = argparse.ArgumentParser() + parser.add_argument( + "--model", required=True, help="Path of model directory.") + return parser.parse_args() + +if __name__ == "__main__": + args = parse_arguments() + model_file = os.path.join(args.model, "inference.pdmodel") + params_file = os.path.join(args.model, "inference.pdiparams") + config_file = os.path.join(args.model, "inference_cls.yaml") + model_buffer = open(model_file, 'rb') + params_buffer = open(params_file, 'rb') + encrypted_model, key = fd.encryption.encrypt(model_buffer.read()) + encrypted_params, key= fd.encryption.encrypt(params_buffer.read(), key) + encrypted_model_dir = args.model + "_encrypt" + model_buffer.close() + params_buffer.close() + os.mkdir(encrypted_model_dir) + with open(os.path.join(encrypted_model_dir, "__model__.encrypted"), "w") as f: + f.write(encrypted_model) + + with open(os.path.join(encrypted_model_dir, "__params__.encrypted"), "w") as f: + f.write(encrypted_params) + + with open(os.path.join(encrypted_model_dir, "encryption_key.txt"), "w") as f: + f.write(key) + print("encryption success") \ No newline at end of file From bf61f84d22177b8fce47036d0fa89723d0666189 Mon Sep 17 00:00:00 2001 From: felixhjh <852142024@qq.com> Date: Mon, 13 Feb 2023 02:35:55 +0000 Subject: [PATCH 19/30] Add some args in encrypt.py --- tutorials/encrypt_model/README.md | 6 ++--- tutorials/encrypt_model/README_CN.md | 4 +-- tutorials/encrypt_model/encrypt.py | 38 +++++++++++++++++++--------- 3 files changed, 31 insertions(+), 17 deletions(-) diff --git a/tutorials/encrypt_model/README.md b/tutorials/encrypt_model/README.md index 8a49c107c..755671686 100644 --- a/tutorials/encrypt_model/README.md +++ b/tutorials/encrypt_model/README.md @@ -6,7 +6,7 @@ This directory provides `encrypt.py` to quickly complete the encryption of the m ## encryption ```bash -# Download deployment example code +# Download deployment example code git clone https://github.com/PaddlePaddle/FastDeploy.git cd FastDeploy/tutorials/encrypt_model @@ -14,7 +14,7 @@ cd FastDeploy/tutorials/encrypt_model wget https://bj.bcebos.com/paddlehub/fastdeploy/ResNet50_vd_infer.tgz tar -xvf ResNet50_vd_infer.tgz -python encrypt.py --model ResNet50_vd_infer +python encrypt.py --model_file ResNet50_vd_infer/inference.pdmodel --params_file ResNet50_vd_infer/inference.pdiparams --encrypted_model_dir ResNet50_vd_infer_encrypt ``` >> **Note** After the encryption is completed, the ResNet50_vd_infer_encrypt folder will be generated, including `__model__.encrypted`, `__params__.encrypted`, `encryption_key.txt` three files, where `encryption_key.txt` contains the encrypted key. At the same time, you need to copy the `inference_cls.yaml` configuration file in the original folder to the ResNet50_vd_infer_encrypt folder for subsequent deployment @@ -43,4 +43,4 @@ option.set_encryption_key(key) fastdeploy::RuntimeOption option; option.SetEncryptionKey(key) ``` ->> **Note** For more details about RuntimeOption, please refer to [RuntimeOption Python Documentation](https://www.paddlepaddle.org.cn/fastdeploy-api-doc/python/html/runtime_option.html), [ RuntimeOption C++ Documentation](https://www.paddlepaddle.org.cn/fastdeploy-api-doc/cpp/html/structfastdeploy_1_1RuntimeOption.html) \ No newline at end of file +>> **Note** For more details about RuntimeOption, please refer to [RuntimeOption Python Documentation](https://www.paddlepaddle.org.cn/fastdeploy-api-doc/python/html/runtime_option.html), [ RuntimeOption C++ Documentation](https://www.paddlepaddle.org.cn/fastdeploy-api-doc/cpp/html/structfastdeploy_1_1RuntimeOption.html) diff --git a/tutorials/encrypt_model/README_CN.md b/tutorials/encrypt_model/README_CN.md index 8230f68e6..c2f80ffd4 100644 --- a/tutorials/encrypt_model/README_CN.md +++ b/tutorials/encrypt_model/README_CN.md @@ -16,7 +16,7 @@ cd FastDeploy/tutorials/encrypt_model wget https://bj.bcebos.com/paddlehub/fastdeploy/ResNet50_vd_infer.tgz tar -xvf ResNet50_vd_infer.tgz -python encrypt.py --model ResNet50_vd_infer +python encrypt.py --model_file ResNet50_vd_infer/inference.pdmodel --params_file ResNet50_vd_infer/inference.pdiparams --encrypted_model_dir ResNet50_vd_infer_encrypt ``` >> **注意** 加密完成后会生成ResNet50_vd_infer_encrypt文件夹，包含`__model__.encrypted`,`__params__.encrypted`,`encryption_key.txt`三个文件，其中`encryption_key.txt`包含加密后的秘钥，同时需要将原文件夹中的、`inference_cls.yaml`配置文件 拷贝至ResNet50_vd_infer_encrypt文件夹，以便后续部署使用 @@ -45,4 +45,4 @@ option.set_encryption_key(key) fastdeploy::RuntimeOption option; option.SetEncryptionKey(key) ``` ->> **注意** RuntimeOption的更多详细信息，请参考[RuntimeOption Python文档](https://www.paddlepaddle.org.cn/fastdeploy-api-doc/python/html/runtime_option.html)，[RuntimeOption C++文档](https://www.paddlepaddle.org.cn/fastdeploy-api-doc/cpp/html/structfastdeploy_1_1RuntimeOption.html) \ No newline at end of file +>> **注意** RuntimeOption的更多详细信息，请参考[RuntimeOption Python文档](https://www.paddlepaddle.org.cn/fastdeploy-api-doc/python/html/runtime_option.html)，[RuntimeOption C++文档](https://www.paddlepaddle.org.cn/fastdeploy-api-doc/cpp/html/structfastdeploy_1_1RuntimeOption.html) diff --git a/tutorials/encrypt_model/encrypt.py b/tutorials/encrypt_model/encrypt.py index 380509042..f4d80ed2f 100644 --- a/tutorials/encrypt_model/encrypt.py +++ b/tutorials/encrypt_model/encrypt.py @@ -1,33 +1,47 @@ import fastdeploy as fd import os + def parse_arguments(): import argparse import ast parser = argparse.ArgumentParser() parser.add_argument( - "--model", required=True, help="Path of model directory.") + "--encrypted_model_dir", + required=False, + help="Path of model directory.") + parser.add_argument( + "--model_file", required=True, help="Path of model file directory.") + parser.add_argument( + "--params_file", + required=True, + help="Path of parameters file directory.") return parser.parse_args() + if __name__ == "__main__": args = parse_arguments() - model_file = os.path.join(args.model, "inference.pdmodel") - params_file = os.path.join(args.model, "inference.pdiparams") - config_file = os.path.join(args.model, "inference_cls.yaml") - model_buffer = open(model_file, 'rb') - params_buffer = open(params_file, 'rb') + model_buffer = open(args.model_file, 'rb') + params_buffer = open(args.params_file, 'rb') encrypted_model, key = fd.encryption.encrypt(model_buffer.read()) - encrypted_params, key= fd.encryption.encrypt(params_buffer.read(), key) - encrypted_model_dir = args.model + "_encrypt" + # use the same key to encrypt parameter file + encrypted_params, key = fd.encryption.encrypt(params_buffer.read(), key) + encrypted_model_dir = "encrypt_model_dir" + if args.encrypted_model_dir: + encrypted_model_dir = args.encrypted_model_dir model_buffer.close() params_buffer.close() os.mkdir(encrypted_model_dir) - with open(os.path.join(encrypted_model_dir, "__model__.encrypted"), "w") as f: + with open(os.path.join(encrypted_model_dir, "__model__.encrypted"), + "w") as f: f.write(encrypted_model) - with open(os.path.join(encrypted_model_dir, "__params__.encrypted"), "w") as f: + with open(os.path.join(encrypted_model_dir, "__params__.encrypted"), + "w") as f: f.write(encrypted_params) - with open(os.path.join(encrypted_model_dir, "encryption_key.txt"), "w") as f: + with open(os.path.join(encrypted_model_dir, "encryption_key.txt"), + "w") as f: f.write(key) - print("encryption success") \ No newline at end of file + print("encryption key: ", key) + print("encryption success") From abfa9fd850025d4c94d45578f72f3aa51124f06e Mon Sep 17 00:00:00 2001 From: wwbitejotunn Date: Mon, 13 Feb 2023 03:11:31 +0000 Subject: [PATCH 20/30] prebind output by shareExternalData --- fastdeploy/pybind/fd_tensor.cc | 93 +++++++++++++------ fastdeploy/pybind/runtime.cc | 1 + .../runtime/backends/paddle/paddle_backend.cc | 38 ++++++-- .../runtime/backends/paddle/paddle_backend.h | 5 +- fastdeploy/runtime/backends/paddle/util.cc | 37 ++++++++ fastdeploy/runtime/runtime.cc | 20 ++++ fastdeploy/runtime/runtime.h | 8 +- python/fastdeploy/runtime.py | 11 ++- 8 files changed, 174 insertions(+), 39 deletions(-) diff --git a/fastdeploy/pybind/fd_tensor.cc b/fastdeploy/pybind/fd_tensor.cc index 0f1d145b3..6e34019f5 100644 --- a/fastdeploy/pybind/fd_tensor.cc +++ b/fastdeploy/pybind/fd_tensor.cc @@ -15,9 +15,9 @@ #include #include "fastdeploy/core/fd_type.h" -#include "fastdeploy/utils/utils.h" #include "fastdeploy/fastdeploy_model.h" #include "fastdeploy/pybind/main.h" +#include "fastdeploy/utils/utils.h" namespace fastdeploy { @@ -68,8 +68,8 @@ DLDataType FDToDlpackType(FDDataType fd_dtype) { break; default: - FDASSERT(false, - "Convert to DlPack, FDType \"%s\" is not supported.", Str(fd_dtype).c_str()); + FDASSERT(false, "Convert to DlPack, FDType \"%s\" is not supported.", + Str(fd_dtype).c_str()); } dl_dtype.code = dl_code; @@ -77,10 +77,8 @@ DLDataType FDToDlpackType(FDDataType fd_dtype) { return dl_dtype; } -FDDataType -DlpackToFDType(const DLDataType& data_type) { - FDASSERT(data_type.lanes == 1, - "FDTensor does not support dlpack lanes != 1") +FDDataType DlpackToFDType(const DLDataType& data_type) { + FDASSERT(data_type.lanes == 1, "FDTensor does not support dlpack lanes != 1") if (data_type.code == DLDataTypeCode::kDLFloat) { if (data_type.bits == 16) { @@ -152,7 +150,7 @@ pybind11::capsule FDTensorToDLPack(FDTensor& fd_tensor) { dlpack_tensor->dl_tensor.dtype = FDToDlpackType(fd_tensor.dtype); dlpack_tensor->dl_tensor.device.device_id = fd_tensor.device_id; - if(fd_tensor.device == Device::GPU) { + if (fd_tensor.device == Device::GPU) { if (fd_tensor.is_pinned_memory) { dlpack_tensor->dl_tensor.device.device_type = DLDeviceType::kDLCUDAHost; } else { @@ -162,8 +160,8 @@ pybind11::capsule FDTensorToDLPack(FDTensor& fd_tensor) { dlpack_tensor->dl_tensor.device.device_type = DLDeviceType::kDLCPU; } - return pybind11::capsule( - static_cast(dlpack_tensor), "dltensor", &DeleteUnusedDltensor); + return pybind11::capsule(static_cast(dlpack_tensor), "dltensor", + &DeleteUnusedDltensor); } FDTensor FDTensorFromDLPack(const std::string& name, @@ -178,9 +176,8 @@ FDTensor FDTensorFromDLPack(const std::string& name, int64_t* strides = dl_managed_tensor->dl_tensor.strides; int ndim = dl_managed_tensor->dl_tensor.ndim; - std::vector dims( - dl_managed_tensor->dl_tensor.shape, - dl_managed_tensor->dl_tensor.shape + ndim); + std::vector dims(dl_managed_tensor->dl_tensor.shape, + dl_managed_tensor->dl_tensor.shape + ndim); // Check if the input is contiguous and in C order if (strides != nullptr) { @@ -196,8 +193,8 @@ FDTensor FDTensorFromDLPack(const std::string& name, } FDASSERT(is_contiguous_c_order, - "DLPack tensor is not contiguous. Only contiguous DLPack " - "tensors that are stored in C-Order are supported."); + "DLPack tensor is not contiguous. Only contiguous DLPack " + "tensors that are stored in C-Order are supported."); } Device device; @@ -216,21 +213,20 @@ FDTensor FDTensorFromDLPack(const std::string& name, is_pinned_memory = true; break; default: - FDASSERT(false, + FDASSERT( + false, ("DLDevice type " + - std::to_string(dl_managed_tensor->dl_tensor.device.device_type) + - " is not support by Python backend.").c_str()); + std::to_string(dl_managed_tensor->dl_tensor.device.device_type) + + " is not support by Python backend.") + .c_str()); break; } - FDDataType dtype = - DlpackToFDType(dl_managed_tensor->dl_tensor.dtype); + FDDataType dtype = DlpackToFDType(dl_managed_tensor->dl_tensor.dtype); PyCapsule_SetName(dlpack_tensor.ptr(), "used_dlpack"); FDTensor fd_tensor(name); - fd_tensor.SetExternalData( - dims, dtype, memory_ptr, device, device_id - ); + fd_tensor.SetExternalData(dims, dtype, memory_ptr, device, device_id); fd_tensor.is_pinned_memory = is_pinned_memory; return fd_tensor; } @@ -242,15 +238,52 @@ void BindFDTensor(pybind11::module& m) { .def_readonly("shape", &FDTensor::shape) .def_readonly("dtype", &FDTensor::dtype) .def_readonly("device", &FDTensor::device) - .def("numpy", [](FDTensor& self) { - return TensorToPyArray(self); - }) + .def("numpy", [](FDTensor& self) { return TensorToPyArray(self); }) .def("data", &FDTensor::MutableData) - .def("from_numpy", [](FDTensor& self, pybind11::array& pyarray, bool share_buffer = false) { - PyArrayToTensor(pyarray, &self, share_buffer); - }) + .def("from_numpy", + [](FDTensor& self, pybind11::array& pyarray, + bool share_buffer = false) { + PyArrayToTensor(pyarray, &self, share_buffer); + }) + .def("from_external_data", + [](const std::string& name, size_t data_addr, + const std::vector& shape, const std::string& data_type, + const std::string& data_place, int device_id) { + auto fd_data_type = FDDataType::UNKNOWN1; + if (data_type == "FP32") { + fd_data_type = FDDataType::FP32; + } else if (data_type == "FP16") { + fd_data_type = FDDataType::FP16; + } else if (data_type == "INT32") { + fd_data_type = FDDataType::INT32; + } else if (data_type == "INT64") { + fd_data_type = FDDataType::INT64; + } else { + FDASSERT(false, + "FDTensor.from_external_data, datatype \"%s\" is not " + "supported.", + data_type.c_str()); + } + + Device fd_data_place; + if (data_place.find("gpu") != data_place.npos) { + fd_data_place = Device::GPU; + } else { + FDASSERT(false, + ("Device type " + data_place + + " is not support by FDTensor.from_external_data.") + .c_str()); + } + void* data_ptr = nullptr; + data_ptr = reinterpret_cast(data_addr); + FDTensor fd_tensor(name); + fd_tensor.SetExternalData(shape, fd_data_type, + static_cast(data_ptr), + fd_data_place, device_id); + return fd_tensor; + }) .def("to_dlpack", &FDTensorToDLPack) - .def("from_dlpack",&FDTensorFromDLPack) + .def("from_dlpack", &FDTensorFromDLPack) .def("print_info", &FDTensor::PrintInfo); } diff --git a/fastdeploy/pybind/runtime.cc b/fastdeploy/pybind/runtime.cc index ca2f4886b..408c3ced2 100644 --- a/fastdeploy/pybind/runtime.cc +++ b/fastdeploy/pybind/runtime.cc @@ -110,6 +110,7 @@ void BindRuntime(pybind11::module& m) { return outputs; }) .def("bind_input_tensor", &Runtime::BindInputTensor) + .def("bind_output_tensor", &Runtime::BindOutputTensor) .def("infer", [](Runtime& self) { self.Infer(); }) .def("get_output_tensor", [](Runtime& self, const std::string& name) { diff --git a/fastdeploy/runtime/backends/paddle/paddle_backend.cc b/fastdeploy/runtime/backends/paddle/paddle_backend.cc index 19493f90b..49cb2532b 100644 --- a/fastdeploy/runtime/backends/paddle/paddle_backend.cc +++ b/fastdeploy/runtime/backends/paddle/paddle_backend.cc @@ -25,6 +25,7 @@ void PaddleBackend::BuildOption(const PaddleBackendOption& option) { if (option.device == Device::GPU) { config_.EnableUseGpu(option.gpu_mem_init_size, option.device_id); if (option_.external_stream_) { + FDINFO << "Will use external stream for Paddle Backend." << std::endl; config_.SetExecStream(option_.external_stream_); } if (option.enable_trt) { @@ -47,7 +48,7 @@ void PaddleBackend::BuildOption(const PaddleBackendOption& option) { config_.SetOptimCacheDir(option.trt_option.serialize_file); } config_.EnableTensorRtEngine(option.trt_option.max_workspace_size, - option.trt_option.max_batch_size, 3, + option.trt_option.max_batch_size, 20, precision, use_static); SetTRTDynamicShapeToConfig(option); } @@ -124,9 +125,10 @@ bool PaddleBackend::InitFromPaddle(const std::string& model_buffer, "file will save to the directory where paddle model saved." << std::endl; use_static = true; + config_.SetOptimCacheDir(option.trt_option.serialize_file); } config_.EnableTensorRtEngine(option.trt_option.max_workspace_size, - option.trt_option.max_batch_size, 3, + option.trt_option.max_batch_size, 20, paddle_infer::PrecisionType::kInt8, use_static, false); SetTRTDynamicShapeToConfig(option); @@ -223,23 +225,47 @@ bool PaddleBackend::Infer(std::vector& inputs, << inputs_desc_.size() << ")." << std::endl; return false; } + // output share backend memory only support CPU or GPU + if (option_.device == Device::IPU) { + copy_to_fd = true; + } RUNTIME_PROFILE_LOOP_H2D_D2H_BEGIN for (size_t i = 0; i < inputs.size(); ++i) { auto handle = predictor_->GetInputHandle(inputs[i].name); ShareTensorFromFDTensor(handle.get(), inputs[i]); } + std::unordered_set prebinded_output_name; + // prebinded output only support for GPU + if (!copy_to_fd) { + for (size_t i = 0; i < (*outputs).size(); ++i) { + auto output_name = (*outputs)[i].name; + // if a output is not prebinded, + // the name of output is expected to be empty. + // We skip here + if (output_name.empty()) { + continue; + } + // Record the prebinded output_name. + // Those outputs do not need PaddleTensorToFDTensor + // after predictor_.Run() + prebinded_output_name.insert(output_name); + auto handle = predictor_->GetOutputHandle(output_name); + ShareOutTensorFromFDTensor(handle.get(), (*outputs)[i]); + } + } RUNTIME_PROFILE_LOOP_BEGIN(1) predictor_->Run(); RUNTIME_PROFILE_LOOP_END - // output share backend memory only support CPU or GPU - if (option_.device == Device::IPU) { - copy_to_fd = true; - } outputs->resize(outputs_desc_.size()); for (size_t i = 0; i < outputs_desc_.size(); ++i) { + // skip prebinded output + if (copy_to_fd == false && + prebinded_output_name.count(outputs_desc_[i].name)) { + continue; + } auto handle = predictor_->GetOutputHandle(outputs_desc_[i].name); if (copy_to_fd) { (*outputs)[i].is_pinned_memory = option_.enable_pinned_memory; diff --git a/fastdeploy/runtime/backends/paddle/paddle_backend.h b/fastdeploy/runtime/backends/paddle/paddle_backend.h index 02c430ade..60079fed6 100755 --- a/fastdeploy/runtime/backends/paddle/paddle_backend.h +++ b/fastdeploy/runtime/backends/paddle/paddle_backend.h @@ -35,6 +35,9 @@ paddle_infer::PlaceType ConvertFDDeviceToPlace(Device device); // Share memory buffer with paddle_infer::Tensor from fastdeploy::FDTensor void ShareTensorFromFDTensor(paddle_infer::Tensor* tensor, FDTensor& fd_tensor); +void ShareOutTensorFromFDTensor(paddle_infer::Tensor* tensor, + FDTensor& fd_tensor); + // convert paddle_infer::Tensor to fastdeploy::FDTensor // if copy_to_fd is true, copy memory data to FDTensor /// else share memory to FDTensor @@ -89,4 +92,4 @@ class PaddleBackend : public BaseBackend { std::vector inputs_desc_; std::vector outputs_desc_; }; -} // namespace fastdeploy \ No newline at end of file +} // namespace fastdeploy diff --git a/fastdeploy/runtime/backends/paddle/util.cc b/fastdeploy/runtime/backends/paddle/util.cc index f117a49bc..bd7ff0944 100644 --- a/fastdeploy/runtime/backends/paddle/util.cc +++ b/fastdeploy/runtime/backends/paddle/util.cc @@ -61,6 +61,43 @@ void ShareTensorFromFDTensor(paddle_infer::Tensor* tensor, Str(fd_tensor.dtype).c_str()); } +void ShareOutTensorFromFDTensor(paddle_infer::Tensor* tensor, + FDTensor& fd_tensor) { + std::vector shape(fd_tensor.shape.begin(), fd_tensor.shape.end()); + auto place = ConvertFDDeviceToPlace(fd_tensor.device); + if (fd_tensor.dtype == FDDataType::FP32) { + if (place == paddle_infer::PlaceType::kGPU) { + tensor->ShareExternalData(static_cast(fd_tensor.MutableData()), + shape, place); + } else { + tensor->CopyToCpu(static_cast(fd_tensor.MutableData())); + } + return; + } else if (fd_tensor.dtype == FDDataType::INT32) { + if (place == paddle_infer::PlaceType::kGPU) { + tensor->ShareExternalData(static_cast(fd_tensor.MutableData()), + shape, place); + } else { + tensor->CopyToCpu(static_cast(fd_tensor.MutableData())); + } + return; + } else if (fd_tensor.dtype == FDDataType::INT64) { + if (place == paddle_infer::PlaceType::kGPU) { + tensor->ShareExternalData(static_cast(fd_tensor.MutableData()), + shape, place); + } else { + tensor->CopyToCpu(static_cast(fd_tensor.MutableData())); + } + return; + } else if (fd_tensor.dtype == FDDataType::UINT8) { + tensor->ShareExternalData(static_cast(fd_tensor.MutableData()), + shape, paddle_infer::PlaceType::kCPU); + return; + } + FDASSERT(false, "Unexpected data type(%s) while infer with PaddleBackend.", + Str(fd_tensor.dtype).c_str()); +} + void PaddleTensorToFDTensor(std::unique_ptr& tensor, FDTensor* fd_tensor, bool copy_to_fd) { auto fd_dtype = PaddleDataTypeToFD(tensor->type()); diff --git a/fastdeploy/runtime/runtime.cc b/fastdeploy/runtime/runtime.cc index 70714e4f0..3f3ccd031 100644 --- a/fastdeploy/runtime/runtime.cc +++ b/fastdeploy/runtime/runtime.cc @@ -198,6 +198,26 @@ void Runtime::BindInputTensor(const std::string& name, FDTensor& input) { } } +void Runtime::BindOutputTensor(const std::string& name, FDTensor& output) { + bool is_exist = false; + for (auto& t : output_tensors_) { + if (t.name == name) { + // FDWARNING << "The output name [" << name << "] is exist." << std::endl; + is_exist = true; + t.SetExternalData(output.shape, output.dtype, output.MutableData(), + output.device, output.device_id); + break; + } + } + if (!is_exist) { + // FDWARNING << "The output name [" << name << "] don't exist." << + // std::endl; + FDTensor new_tensor(name); + new_tensor.SetExternalData(output.shape, output.dtype, output.MutableData(), + output.device, output.device_id); + output_tensors_.emplace_back(std::move(new_tensor)); + } +} FDTensor* Runtime::GetOutputTensor(const std::string& name) { for (auto& t : output_tensors_) { if (t.name == name) { diff --git a/fastdeploy/runtime/runtime.h b/fastdeploy/runtime/runtime.h index 6e7dc9629..66ce9c94e 100755 --- a/fastdeploy/runtime/runtime.h +++ b/fastdeploy/runtime/runtime.h @@ -72,6 +72,12 @@ struct FASTDEPLOY_DECL Runtime { /** \brief Bind FDTensor by name, no copy and share input memory */ void BindInputTensor(const std::string& name, FDTensor& input); + + /** \brief Bind FDTensor by name, no copy and share output memory. + * Please make share the correctness of tensor shape of output. + */ + void BindOutputTensor(const std::string& name, FDTensor& output); + /** \brief Get output FDTensor by name, no copy and share backend output memory */ FDTensor* GetOutputTensor(const std::string& name); @@ -99,7 +105,7 @@ struct FASTDEPLOY_DECL Runtime { */ double GetProfileTime() { return backend_->benchmark_result_.time_of_runtime; - } + } private: void CreateOrtBackend(); diff --git a/python/fastdeploy/runtime.py b/python/fastdeploy/runtime.py index 47659c98c..1251b955c 100644 --- a/python/fastdeploy/runtime.py +++ b/python/fastdeploy/runtime.py @@ -72,6 +72,14 @@ class Runtime: """ self._runtime.bind_input_tensor(name, fdtensor) + def bind_output_tensor(self, name, fdtensor): + """Bind FDTensor by name, no copy and share output memory + + :param name: (str)The name of output data. + :param fdtensor: (fastdeploy.FDTensor)The output FDTensor. + """ + self._runtime.bind_output_tensor(name, fdtensor) + def zero_copy_infer(self): """No params inference the model. @@ -656,7 +664,8 @@ class RuntimeOption: continue if hasattr(getattr(self._option, attr), "__call__"): continue - message += " {} : {}\t\n".format(attr, getattr(self._option, attr)) + message += " {} : {}\t\n".format(attr, + getattr(self._option, attr)) message.strip("\n") message += ")" return message From 45b070e4959c82e0960b83697ce9260a89951832 Mon Sep 17 00:00:00 2001 From: felixhjh <852142024@qq.com> Date: Mon, 13 Feb 2023 06:31:04 +0000 Subject: [PATCH 21/30] Format code --- fastdeploy/runtime/runtime.cc | 46 +++++++++++++++------------- fastdeploy/runtime/runtime.h | 4 +-- fastdeploy/runtime/runtime_option.cc | 12 ++++---- fastdeploy/runtime/runtime_option.h | 2 +- python/fastdeploy/runtime.py | 16 +++++----- 5 files changed, 42 insertions(+), 38 deletions(-) diff --git a/fastdeploy/runtime/runtime.cc b/fastdeploy/runtime/runtime.cc index 2a00dfda4..67774a306 100644 --- a/fastdeploy/runtime/runtime.cc +++ b/fastdeploy/runtime/runtime.cc @@ -104,30 +104,32 @@ bool AutoSelectBackend(RuntimeOption& option) { bool Runtime::Init(const RuntimeOption& _option) { option = _option; + // decrypt encrypted model if ("" != option.encryption_key_) { - #ifdef ENABLE_ENCRYPTION - if (option.model_from_memory_) { - option.model_file = Decrypt(option.model_file, option.encryption_key_); - if (!(option.params_file.empty())) { - option.params_file = Decrypt(option.params_file, option.encryption_key_); - } - } else { - std::string model_buffer = ""; - FDASSERT(ReadBinaryFromFile(option.model_file, &model_buffer), - "Fail to read binary from model file"); - option.model_file = Decrypt(model_buffer, option.encryption_key_); - if (!(option.params_file.empty())) { - std::string params_buffer = ""; - FDASSERT(ReadBinaryFromFile(option.params_file, ¶ms_buffer), - "Fail to read binary from parameter file"); - option.params_file = Decrypt(params_buffer, option.encryption_key_); - } - option.model_from_memory_ = true; +#ifdef ENABLE_ENCRYPTION + if (option.model_from_memory_) { + option.model_file = Decrypt(option.model_file, option.encryption_key_); + if (!(option.params_file.empty())) { + option.params_file = + Decrypt(option.params_file, option.encryption_key_); } - #else - FDERROR << "The FastDeploy didn't compile with encryption function." - << std::endl; - #endif + } else { + std::string model_buffer = ""; + FDASSERT(ReadBinaryFromFile(option.model_file, &model_buffer), + "Fail to read binary from model file"); + option.model_file = Decrypt(model_buffer, option.encryption_key_); + if (!(option.params_file.empty())) { + std::string params_buffer = ""; + FDASSERT(ReadBinaryFromFile(option.params_file, ¶ms_buffer), + "Fail to read binary from parameter file"); + option.params_file = Decrypt(params_buffer, option.encryption_key_); + } + option.model_from_memory_ = true; + } +#else + FDERROR << "The FastDeploy didn't compile with encryption function." + << std::endl; +#endif } // Choose default backend by model format and device if backend is not // specified diff --git a/fastdeploy/runtime/runtime.h b/fastdeploy/runtime/runtime.h index 772773007..fa8b8f198 100755 --- a/fastdeploy/runtime/runtime.h +++ b/fastdeploy/runtime/runtime.h @@ -24,7 +24,7 @@ #include "fastdeploy/runtime/runtime_option.h" #include "fastdeploy/utils/perf.h" #ifdef ENABLE_ENCRYPTION - #include "fastdeploy/encryption/include/decrypt.h" +#include "fastdeploy/encryption/include/decrypt.h" #endif /** \brief All C++ FastDeploy APIs are defined inside this namespace @@ -102,7 +102,7 @@ struct FASTDEPLOY_DECL Runtime { */ double GetProfileTime() { return backend_->benchmark_result_.time_of_runtime; - } + } private: void CreateOrtBackend(); diff --git a/fastdeploy/runtime/runtime_option.cc b/fastdeploy/runtime/runtime_option.cc index 8568b3b7f..d074a9603 100644 --- a/fastdeploy/runtime/runtime_option.cc +++ b/fastdeploy/runtime/runtime_option.cc @@ -37,12 +37,12 @@ void RuntimeOption::SetModelBuffer(const std::string& model_buffer, } void RuntimeOption::SetEncryptionKey(const std::string& encryption_key) { - #ifdef ENABLE_ENCRYPTION - encryption_key_ = encryption_key; - #else - FDERROR << "The FastDeploy didn't compile with encryption function." - << std::endl; - #endif +#ifdef ENABLE_ENCRYPTION + encryption_key_ = encryption_key; +#else + FDERROR << "The FastDeploy didn't compile with encryption function." + << std::endl; +#endif } void RuntimeOption::UseGpu(int gpu_id) { diff --git a/fastdeploy/runtime/runtime_option.h b/fastdeploy/runtime/runtime_option.h index b263f4ad3..a10ed9845 100644 --- a/fastdeploy/runtime/runtime_option.h +++ b/fastdeploy/runtime/runtime_option.h @@ -207,7 +207,7 @@ struct FASTDEPLOY_DECL RuntimeOption { // *** The belowing api are deprecated, will be removed in v1.2.0 // *** Do not use it anymore - void SetPaddleMKLDNN(bool pd_mkldnn = true); + void SetPaddleMKLDNN(bool pd_mkldnn = true); void EnablePaddleToTrt(); void DeletePaddleBackendPass(const std::string& delete_pass_name); void EnablePaddleLogInfo(); diff --git a/python/fastdeploy/runtime.py b/python/fastdeploy/runtime.py index 1d2fc1c1d..a9004a15a 100644 --- a/python/fastdeploy/runtime.py +++ b/python/fastdeploy/runtime.py @@ -187,8 +187,7 @@ class RuntimeOption: return self._option.set_model_buffer(model_buffer, params_buffer, model_format) - def set_encryption_key(self, - encryption_key): + def set_encryption_key(self, encryption_key): """When loading encrypted model, encryption_key is required to decrypte model :param encryption_key: (str)The key for decrypting model """ @@ -590,10 +589,12 @@ class RuntimeOption: replica_num=1, available_memory_proportion=1.0, enable_half_partial=False): - logging.warning("`RuntimeOption.set_ipu_config` will be deprecated in v1.2.0, please use `RuntimeOption.paddle_infer_option.set_ipu_config()` instead.") - self._option.paddle_infer_option.set_ipu_config(enable_fp16, replica_num, - available_memory_proportion, - enable_half_partial) + logging.warning( + "`RuntimeOption.set_ipu_config` will be deprecated in v1.2.0, please use `RuntimeOption.paddle_infer_option.set_ipu_config()` instead." + ) + self._option.paddle_infer_option.set_ipu_config( + enable_fp16, replica_num, available_memory_proportion, + enable_half_partial) @property def poros_option(self): @@ -664,7 +665,8 @@ class RuntimeOption: continue if hasattr(getattr(self._option, attr), "__call__"): continue - message += " {} : {}\t\n".format(attr, getattr(self._option, attr)) + message += " {} : {}\t\n".format(attr, + getattr(self._option, attr)) message.strip("\n") message += ")" return message From 47b1d27fbb11e88dd4ccfe13d11e4561c1d9eb4e Mon Sep 17 00:00:00 2001 From: WJJ1995 Date: Mon, 13 Feb 2023 16:12:54 +0800 Subject: [PATCH 22/30] [Benchmark] Add macros for benchmark (#1301) * add GPL lisence * add GPL-3.0 lisence * add GPL-3.0 lisence * add GPL-3.0 lisence * support yolov8 * add pybind for yolov8 * add yolov8 readme * add cpp benchmark * add cpu and gpu mem * public part split * add runtime mode * fixed bugs * add cpu_thread_nums * deal with comments * deal with comments * deal with comments * rm useless code * add FASTDEPLOY_DECL * add FASTDEPLOY_DECL * fixed for windows * mv rss to pss * mv rss to pss * Update utils.cc * use thread to collect mem * Add ResourceUsageMonitor * rm useless code * fixed bug * fixed typo * update ResourceUsageMonitor * fixed bug * fixed bug * add note for ResourceUsageMonitor * deal with comments * add macros * deal with comments * deal with comments * deal with comments * re-lint --------- Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> --- benchmark/cpp/benchmark_ppyolov8.cc | 110 ++++------------------------ benchmark/cpp/benchmark_yolov5.cc | 97 ++++-------------------- benchmark/cpp/flags.h | 73 ------------------ benchmark/cpp/macros.h | 70 ++++++++++++++++++ benchmark/cpp/option.h | 92 +++++++++++++++++++++++ 5 files changed, 190 insertions(+), 252 deletions(-) mode change 100755 => 100644 benchmark/cpp/benchmark_yolov5.cc create mode 100755 benchmark/cpp/macros.h create mode 100755 benchmark/cpp/option.h diff --git a/benchmark/cpp/benchmark_ppyolov8.cc b/benchmark/cpp/benchmark_ppyolov8.cc index cff374200..545474635 100644 --- a/benchmark/cpp/benchmark_ppyolov8.cc +++ b/benchmark/cpp/benchmark_ppyolov8.cc @@ -12,9 +12,9 @@ // See the License for the specific language governing permissions and // limitations under the License. -#include "fastdeploy/benchmark/utils.h" -#include "fastdeploy/vision.h" #include "flags.h" +#include "macros.h" +#include "option.h" #ifdef WIN32 const char sep = '\\'; @@ -22,104 +22,24 @@ const char sep = '\\'; const char sep = '/'; #endif -bool RunModel(std::string model_dir, std::string image_file, size_t warmup, - size_t repeats, size_t dump_period, std::string cpu_mem_file_name, - std::string gpu_mem_file_name) { +int main(int argc, char* argv[]) { + google::ParseCommandLineFlags(&argc, &argv, true); + auto im = cv::imread(FLAGS_image); // Initialization auto option = fastdeploy::RuntimeOption(); if (!CreateRuntimeOption(&option)) { PrintUsage(); return false; } - auto model_file = model_dir + sep + "model.pdmodel"; - auto params_file = model_dir + sep + "model.pdiparams"; - auto config_file = model_dir + sep + "infer_cfg.yml"; - - if (FLAGS_profile_mode == "runtime") { - option.EnableProfiling(FLAGS_include_h2d_d2h, repeats, warmup); - } - auto model = fastdeploy::vision::detection::PaddleYOLOv8( + auto model_file = FLAGS_model + sep + "model.pdmodel"; + auto params_file = FLAGS_model + sep + "model.pdiparams"; + auto config_file = FLAGS_model + sep + "infer_cfg.yml"; + auto model_ppyolov8 = fastdeploy::vision::detection::PaddleYOLOv8( model_file, params_file, config_file, option); - if (!model.Initialized()) { - std::cerr << "Failed to initialize." << std::endl; - return false; - } - auto im = cv::imread(image_file); - // For Runtime - if (FLAGS_profile_mode == "runtime") { - fastdeploy::vision::DetectionResult res; - if (!model.Predict(im, &res)) { - std::cerr << "Failed to predict." << std::endl; - return false; - } - double profile_time = model.GetProfileTime() * 1000; - std::cout << "Runtime(ms): " << profile_time << "ms." << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im, res); - cv::imwrite("vis_result.jpg", vis_im); - std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; - } else { - // For End2End - // Step1: warm up for warmup times - std::cout << "Warmup " << warmup << " times..." << std::endl; - for (int i = 0; i < warmup; i++) { - fastdeploy::vision::DetectionResult res; - if (!model.Predict(im, &res)) { - std::cerr << "Failed to predict." << std::endl; - return false; - } - } - std::vector end2end_statis; - // Step2: repeat for repeats times - std::cout << "Counting time..." << std::endl; - fastdeploy::TimeCounter tc; - fastdeploy::vision::DetectionResult res; - for (int i = 0; i < repeats; i++) { - if (FLAGS_collect_memory_info && i % dump_period == 0) { - fastdeploy::benchmark::DumpCurrentCpuMemoryUsage(cpu_mem_file_name); -#if defined(WITH_GPU) - fastdeploy::benchmark::DumpCurrentGpuMemoryUsage(gpu_mem_file_name, - FLAGS_device_id); -#endif - } - tc.Start(); - if (!model.Predict(im, &res)) { - std::cerr << "Failed to predict." << std::endl; - return false; - } - tc.End(); - end2end_statis.push_back(tc.Duration() * 1000); - } - float end2end = std::accumulate(end2end_statis.end() - repeats, - end2end_statis.end(), 0.f) / - repeats; - std::cout << "End2End(ms): " << end2end << "ms." << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im, res); - cv::imwrite("vis_result.jpg", vis_im); - std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; - } - - return true; -} - -int main(int argc, char* argv[]) { - google::ParseCommandLineFlags(&argc, &argv, true); - int repeats = FLAGS_repeat; - int warmup = FLAGS_warmup; - int dump_period = FLAGS_dump_period; - std::string cpu_mem_file_name = "result_cpu.txt"; - std::string gpu_mem_file_name = "result_gpu.txt"; - // Run model - if (RunModel(FLAGS_model, FLAGS_image, warmup, repeats, dump_period, - cpu_mem_file_name, gpu_mem_file_name) != true) { - exit(1); - } - if (FLAGS_collect_memory_info) { - float cpu_mem = fastdeploy::benchmark::GetCpuMemoryUsage(cpu_mem_file_name); - std::cout << "cpu_pss_mb: " << cpu_mem << "MB." << std::endl; -#if defined(WITH_GPU) - float gpu_mem = fastdeploy::benchmark::GetGpuMemoryUsage(gpu_mem_file_name); - std::cout << "gpu_pss_mb: " << gpu_mem << "MB." << std::endl; -#endif - } + fastdeploy::vision::DetectionResult res; + BENCHMARK_MODEL(model_ppyolov8, model_ppyolov8.Predict(im, &res)) + auto vis_im = fastdeploy::vision::VisDetection(im, res); + cv::imwrite("vis_result.jpg", vis_im); + std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; return 0; -} +} \ No newline at end of file diff --git a/benchmark/cpp/benchmark_yolov5.cc b/benchmark/cpp/benchmark_yolov5.cc old mode 100755 new mode 100644 index 2e5df6b1c..5b2cab855 --- a/benchmark/cpp/benchmark_yolov5.cc +++ b/benchmark/cpp/benchmark_yolov5.cc @@ -12,96 +12,25 @@ // See the License for the specific language governing permissions and // limitations under the License. -#include "fastdeploy/benchmark/utils.h" -#include "fastdeploy/vision.h" #include "flags.h" +#include "macros.h" +#include "option.h" -bool RunModel(std::string model_file, std::string image_file, size_t warmup, - size_t repeats, size_t sampling_interval) { +int main(int argc, char* argv[]) { + google::ParseCommandLineFlags(&argc, &argv, true); + auto im = cv::imread(FLAGS_image); // Initialization auto option = fastdeploy::RuntimeOption(); if (!CreateRuntimeOption(&option)) { PrintUsage(); return false; } - if (FLAGS_profile_mode == "runtime") { - option.EnableProfiling(FLAGS_include_h2d_d2h, repeats, warmup); - } - auto model = fastdeploy::vision::detection::YOLOv5(model_file, "", option); - if (!model.Initialized()) { - std::cerr << "Failed to initialize." << std::endl; - return false; - } - auto im = cv::imread(image_file); - // For collect memory info - fastdeploy::benchmark::ResourceUsageMonitor resource_moniter( - sampling_interval, FLAGS_device_id); - if (FLAGS_collect_memory_info) { - resource_moniter.Start(); - } - // For Runtime - if (FLAGS_profile_mode == "runtime") { - fastdeploy::vision::DetectionResult res; - if (!model.Predict(im, &res)) { - std::cerr << "Failed to predict." << std::endl; - return false; - } - double profile_time = model.GetProfileTime() * 1000; - std::cout << "Runtime(ms): " << profile_time << "ms." << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im, res); - cv::imwrite("vis_result.jpg", vis_im); - std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; - } else { - // For End2End - // Step1: warm up for warmup times - std::cout << "Warmup " << warmup << " times..." << std::endl; - for (int i = 0; i < warmup; i++) { - fastdeploy::vision::DetectionResult res; - if (!model.Predict(im, &res)) { - std::cerr << "Failed to predict." << std::endl; - return false; - } - } - // Step2: repeat for repeats times - std::cout << "Counting time..." << std::endl; - std::cout << "Repeat " << repeats << " times..." << std::endl; - fastdeploy::vision::DetectionResult res; - fastdeploy::TimeCounter tc; - tc.Start(); - for (int i = 0; i < repeats; i++) { - if (!model.Predict(im, &res)) { - std::cerr << "Failed to predict." << std::endl; - return false; - } - } - tc.End(); - double end2end = tc.Duration() / repeats * 1000; - std::cout << "End2End(ms): " << end2end << "ms." << std::endl; - auto vis_im = fastdeploy::vision::VisDetection(im, res); - cv::imwrite("vis_result.jpg", vis_im); - std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; - } - if (FLAGS_collect_memory_info) { - float cpu_mem = resource_moniter.GetMaxCpuMem(); - float gpu_mem = resource_moniter.GetMaxGpuMem(); - float gpu_util = resource_moniter.GetMaxGpuUtil(); - std::cout << "cpu_pss_mb: " << cpu_mem << "MB." << std::endl; - std::cout << "gpu_pss_mb: " << gpu_mem << "MB." << std::endl; - std::cout << "gpu_util: " << gpu_util << std::endl; - resource_moniter.Stop(); - } - - return true; -} - -int main(int argc, char* argv[]) { - google::ParseCommandLineFlags(&argc, &argv, true); - int repeats = FLAGS_repeat; - int warmup = FLAGS_warmup; - int sampling_interval = FLAGS_sampling_interval; - // Run model - if (!RunModel(FLAGS_model, FLAGS_image, warmup, repeats, sampling_interval)) { - exit(1); - } + auto model_yolov5 = + fastdeploy::vision::detection::YOLOv5(FLAGS_model, "", option); + fastdeploy::vision::DetectionResult res; + BENCHMARK_MODEL(model_yolov5, model_yolov5.Predict(im, &res)) + auto vis_im = fastdeploy::vision::VisDetection(im, res); + cv::imwrite("vis_result.jpg", vis_im); + std::cout << "Visualized result saved in ./vis_result.jpg" << std::endl; return 0; -} \ No newline at end of file +} diff --git a/benchmark/cpp/flags.h b/benchmark/cpp/flags.h index 64f22c702..7f8c3a29f 100755 --- a/benchmark/cpp/flags.h +++ b/benchmark/cpp/flags.h @@ -15,7 +15,6 @@ #pragma once #include "gflags/gflags.h" -#include "fastdeploy/utils/perf.h" DEFINE_string(model, "", "Directory of the inference model."); DEFINE_string(image, "", "Path of the image file."); @@ -49,75 +48,3 @@ void PrintUsage() { std::cout << "Default value of backend: default" << std::endl; std::cout << "Default value of use_fp16: false" << std::endl; } - -bool CreateRuntimeOption(fastdeploy::RuntimeOption* option) { - if (FLAGS_device == "gpu") { - option->UseGpu(FLAGS_device_id); - if (FLAGS_backend == "ort") { - option->UseOrtBackend(); - } else if (FLAGS_backend == "paddle") { - option->UsePaddleInferBackend(); - } else if (FLAGS_backend == "trt" || FLAGS_backend == "paddle_trt") { - option->UseTrtBackend(); - if (FLAGS_backend == "paddle_trt") { - option->EnablePaddleToTrt(); - } - if (FLAGS_use_fp16) { - option->EnableTrtFP16(); - } - } else if (FLAGS_backend == "default") { - return true; - } else { - std::cout << "While inference with GPU, only support " - "default/ort/paddle/trt/paddle_trt now, " - << FLAGS_backend << " is not supported." << std::endl; - return false; - } - } else if (FLAGS_device == "cpu") { - option->SetCpuThreadNum(FLAGS_cpu_thread_nums); - if (FLAGS_backend == "ort") { - option->UseOrtBackend(); - } else if (FLAGS_backend == "ov") { - option->UseOpenVINOBackend(); - } else if (FLAGS_backend == "paddle") { - option->UsePaddleInferBackend(); - } else if (FLAGS_backend == "lite") { - option->UsePaddleLiteBackend(); - if (FLAGS_use_fp16) { - option->EnableLiteFP16(); - } - } else if (FLAGS_backend == "default") { - return true; - } else { - std::cout << "While inference with CPU, only support " - "default/ort/ov/paddle/lite now, " - << FLAGS_backend << " is not supported." << std::endl; - return false; - } - } else if (FLAGS_device == "xpu") { - option->UseKunlunXin(FLAGS_device_id); - if (FLAGS_backend == "ort") { - option->UseOrtBackend(); - } else if (FLAGS_backend == "paddle") { - option->UsePaddleInferBackend(); - } else if (FLAGS_backend == "lite") { - option->UsePaddleLiteBackend(); - if (FLAGS_use_fp16) { - option->EnableLiteFP16(); - } - } else if (FLAGS_backend == "default") { - return true; - } else { - std::cout << "While inference with XPU, only support " - "default/ort/paddle/lite now, " - << FLAGS_backend << " is not supported." << std::endl; - return false; - } - } else { - std::cerr << "Only support device CPU/GPU/XPU now, " << FLAGS_device - << " is not supported." << std::endl; - return false; - } - - return true; -} diff --git a/benchmark/cpp/macros.h b/benchmark/cpp/macros.h new file mode 100755 index 000000000..bebd26e0d --- /dev/null +++ b/benchmark/cpp/macros.h @@ -0,0 +1,70 @@ +// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. +#pragma once + +#include "fastdeploy/benchmark/utils.h" +#include "fastdeploy/utils/perf.h" + +#define BENCHMARK_MODEL(MODEL_NAME, BENCHMARK_FUNC) \ +{ \ + std::cout << "====" << #MODEL_NAME << "====" << std::endl; \ + if (!MODEL_NAME.Initialized()) { \ + std::cerr << "Failed to initialize." << std::endl; \ + return 0; \ + } \ + auto __im__ = cv::imread(FLAGS_image); \ + fastdeploy::benchmark::ResourceUsageMonitor __resource_moniter__( \ + FLAGS_sampling_interval, FLAGS_device_id); \ + if (FLAGS_collect_memory_info) { \ + __resource_moniter__.Start(); \ + } \ + if (FLAGS_profile_mode == "runtime") { \ + if (!BENCHMARK_FUNC) { \ + std::cerr << "Failed to predict." << std::endl; \ + return 0; \ + } \ + double __profile_time__ = MODEL_NAME.GetProfileTime() * 1000; \ + std::cout << "Runtime(ms): " << __profile_time__ << "ms." << std::endl; \ + } else { \ + std::cout << "Warmup " << FLAGS_warmup << " times..." << std::endl; \ + for (int __i__ = 0; __i__ < FLAGS_warmup; __i__++) { \ + if (!BENCHMARK_FUNC) { \ + std::cerr << "Failed to predict." << std::endl; \ + return 0; \ + } \ + } \ + std::cout << "Counting time..." << std::endl; \ + std::cout << "Repeat " << FLAGS_repeat << " times..." << std::endl; \ + fastdeploy::TimeCounter __tc__; \ + __tc__.Start(); \ + for (int __i__ = 0; __i__ < FLAGS_repeat; __i__++) { \ + if (!BENCHMARK_FUNC) { \ + std::cerr << "Failed to predict." << std::endl; \ + return 0; \ + } \ + } \ + __tc__.End(); \ + double __end2end__ = __tc__.Duration() / FLAGS_repeat * 1000; \ + std::cout << "End2End(ms): " << __end2end__ << "ms." << std::endl; \ + } \ + if (FLAGS_collect_memory_info) { \ + float __cpu_mem__ = __resource_moniter__.GetMaxCpuMem(); \ + float __gpu_mem__ = __resource_moniter__.GetMaxGpuMem(); \ + float __gpu_util__ = __resource_moniter__.GetMaxGpuUtil(); \ + std::cout << "cpu_pss_mb: " << __cpu_mem__ << "MB." << std::endl; \ + std::cout << "gpu_pss_mb: " << __gpu_mem__ << "MB." << std::endl; \ + std::cout << "gpu_util: " << __gpu_util__ << std::endl; \ + __resource_moniter__.Stop(); \ + } \ +} diff --git a/benchmark/cpp/option.h b/benchmark/cpp/option.h new file mode 100755 index 000000000..9989255e5 --- /dev/null +++ b/benchmark/cpp/option.h @@ -0,0 +1,92 @@ +// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once + +#include "fastdeploy/vision.h" + +static bool CreateRuntimeOption(fastdeploy::RuntimeOption* option) { + if (FLAGS_profile_mode == "runtime") { + option->EnableProfiling(FLAGS_include_h2d_d2h, FLAGS_repeat, FLAGS_warmup); + } + if (FLAGS_device == "gpu") { + option->UseGpu(FLAGS_device_id); + if (FLAGS_backend == "ort") { + option->UseOrtBackend(); + } else if (FLAGS_backend == "paddle") { + option->UsePaddleInferBackend(); + } else if (FLAGS_backend == "trt" || FLAGS_backend == "paddle_trt") { + option->UseTrtBackend(); + if (FLAGS_backend == "paddle_trt") { + option->EnablePaddleToTrt(); + } + if (FLAGS_use_fp16) { + option->EnableTrtFP16(); + } + } else if (FLAGS_backend == "default") { + return true; + } else { + std::cout << "While inference with GPU, only support " + "default/ort/paddle/trt/paddle_trt now, " + << FLAGS_backend << " is not supported." << std::endl; + return false; + } + } else if (FLAGS_device == "cpu") { + option->SetCpuThreadNum(FLAGS_cpu_thread_nums); + if (FLAGS_backend == "ort") { + option->UseOrtBackend(); + } else if (FLAGS_backend == "ov") { + option->UseOpenVINOBackend(); + } else if (FLAGS_backend == "paddle") { + option->UsePaddleInferBackend(); + } else if (FLAGS_backend == "lite") { + option->UsePaddleLiteBackend(); + if (FLAGS_use_fp16) { + option->EnableLiteFP16(); + } + } else if (FLAGS_backend == "default") { + return true; + } else { + std::cout << "While inference with CPU, only support " + "default/ort/ov/paddle/lite now, " + << FLAGS_backend << " is not supported." << std::endl; + return false; + } + } else if (FLAGS_device == "xpu") { + option->UseKunlunXin(FLAGS_device_id); + if (FLAGS_backend == "ort") { + option->UseOrtBackend(); + } else if (FLAGS_backend == "paddle") { + option->UsePaddleInferBackend(); + } else if (FLAGS_backend == "lite") { + option->UsePaddleLiteBackend(); + if (FLAGS_use_fp16) { + option->EnableLiteFP16(); + } + } else if (FLAGS_backend == "default") { + return true; + } else { + std::cout << "While inference with XPU, only support " + "default/ort/paddle/lite now, " + << FLAGS_backend << " is not supported." << std::endl; + return false; + } + } else { + std::cerr << "Only support device CPU/GPU/XPU now, " << FLAGS_device + << " is not supported." << std::endl; + return false; + } + + return true; +} From f1ab47a4efdcfcd5ad467565c7881323d22e4175 Mon Sep 17 00:00:00 2001 From: wwbitejotunn Date: Mon, 13 Feb 2023 18:16:45 +0000 Subject: [PATCH 23/30] code revine --- fastdeploy/runtime/backends/paddle/paddle_backend.cc | 5 ++--- fastdeploy/runtime/runtime.cc | 5 ++--- 2 files changed, 4 insertions(+), 6 deletions(-) diff --git a/fastdeploy/runtime/backends/paddle/paddle_backend.cc b/fastdeploy/runtime/backends/paddle/paddle_backend.cc index 73f0df673..09dbe812a 100644 --- a/fastdeploy/runtime/backends/paddle/paddle_backend.cc +++ b/fastdeploy/runtime/backends/paddle/paddle_backend.cc @@ -51,7 +51,7 @@ void PaddleBackend::BuildOption(const PaddleBackendOption& option) { config_.SetOptimCacheDir(opt_cache_dir); } config_.EnableTensorRtEngine(option.trt_option.max_workspace_size, - option.trt_option.max_batch_size, 20, + option.trt_option.max_batch_size, 3, precision, use_static); SetTRTDynamicShapeToConfig(option); } @@ -128,10 +128,9 @@ bool PaddleBackend::InitFromPaddle(const std::string& model_buffer, "file will save to the directory where paddle model saved." << std::endl; use_static = true; - config_.SetOptimCacheDir(option.trt_option.serialize_file); } config_.EnableTensorRtEngine(option.trt_option.max_workspace_size, - option.trt_option.max_batch_size, 20, + option.trt_option.max_batch_size, 3, paddle_infer::PrecisionType::kInt8, use_static, false); SetTRTDynamicShapeToConfig(option); diff --git a/fastdeploy/runtime/runtime.cc b/fastdeploy/runtime/runtime.cc index 3f3ccd031..160894a68 100644 --- a/fastdeploy/runtime/runtime.cc +++ b/fastdeploy/runtime/runtime.cc @@ -202,7 +202,7 @@ void Runtime::BindOutputTensor(const std::string& name, FDTensor& output) { bool is_exist = false; for (auto& t : output_tensors_) { if (t.name == name) { - // FDWARNING << "The output name [" << name << "] is exist." << std::endl; + FDINFO << "The output name [" << name << "] is exist." << std::endl; is_exist = true; t.SetExternalData(output.shape, output.dtype, output.MutableData(), output.device, output.device_id); @@ -210,8 +210,7 @@ void Runtime::BindOutputTensor(const std::string& name, FDTensor& output) { } } if (!is_exist) { - // FDWARNING << "The output name [" << name << "] don't exist." << - // std::endl; + FDINFO << "The output name [" << name << "] is prebinded added into output tensor list." << std::endl; FDTensor new_tensor(name); new_tensor.SetExternalData(output.shape, output.dtype, output.MutableData(), output.device, output.device_id); From 727fdc58636e2fdbfdb64bd88685307c8aad5f27 Mon Sep 17 00:00:00 2001 From: wwbitejotunn Date: Tue, 14 Feb 2023 06:22:47 +0000 Subject: [PATCH 24/30] reuse setStream c++ interface, add set_raw_stream python api --- fastdeploy/runtime/option_pybind.cc | 5 ++++- fastdeploy/runtime/runtime_option.cc | 4 ---- fastdeploy/runtime/runtime_option.h | 2 -- python/fastdeploy/runtime.py | 15 +++++++++++---- 4 files changed, 15 insertions(+), 11 deletions(-) diff --git a/fastdeploy/runtime/option_pybind.cc b/fastdeploy/runtime/option_pybind.cc index d5a6187c8..edec738f7 100644 --- a/fastdeploy/runtime/option_pybind.cc +++ b/fastdeploy/runtime/option_pybind.cc @@ -48,7 +48,10 @@ void BindOption(pybind11::module& m) { .def_readwrite("poros_option", &RuntimeOption::poros_option) .def_readwrite("paddle_infer_option", &RuntimeOption::paddle_infer_option) .def("set_external_stream", &RuntimeOption::SetExternalStream) - .def("set_external_raw_stream", &RuntimeOption::SetExternalRawStream) + .def("set_external_raw_stream", + [](RuntimeOption& self, size_t external_stream) { + self.SetExternalStream(reinterpret_cast(external_stream)); + }) .def("set_cpu_thread_num", &RuntimeOption::SetCpuThreadNum) .def("use_paddle_backend", &RuntimeOption::UsePaddleBackend) .def("use_poros_backend", &RuntimeOption::UsePorosBackend) diff --git a/fastdeploy/runtime/runtime_option.cc b/fastdeploy/runtime/runtime_option.cc index 8a57c33f7..c09352d58 100644 --- a/fastdeploy/runtime/runtime_option.cc +++ b/fastdeploy/runtime/runtime_option.cc @@ -93,10 +93,6 @@ void RuntimeOption::SetExternalStream(void* external_stream) { external_stream_ = external_stream; } -void RuntimeOption::SetExternalRawStream(size_t external_stream) { - external_stream_ = (void*) external_stream; -} - void RuntimeOption::SetCpuThreadNum(int thread_num) { FDASSERT(thread_num > 0, "The thread_num must be greater than 0."); cpu_thread_num = thread_num; diff --git a/fastdeploy/runtime/runtime_option.h b/fastdeploy/runtime/runtime_option.h index 84cfebc84..eb26cbfaa 100755 --- a/fastdeploy/runtime/runtime_option.h +++ b/fastdeploy/runtime/runtime_option.h @@ -105,8 +105,6 @@ struct FASTDEPLOY_DECL RuntimeOption { void SetExternalStream(void* external_stream); - void SetExternalRawStream(size_t external_stream); - /* * @brief Set number of cpu threads while inference on CPU, by default it will decided by the different backends */ diff --git a/python/fastdeploy/runtime.py b/python/fastdeploy/runtime.py index 20b2d8ad9..4562d2f8d 100644 --- a/python/fastdeploy/runtime.py +++ b/python/fastdeploy/runtime.py @@ -591,10 +591,12 @@ class RuntimeOption: replica_num=1, available_memory_proportion=1.0, enable_half_partial=False): - logging.warning("`RuntimeOption.set_ipu_config` will be deprecated in v1.2.0, please use `RuntimeOption.paddle_infer_option.set_ipu_config()` instead.") - self._option.paddle_infer_option.set_ipu_config(enable_fp16, replica_num, - available_memory_proportion, - enable_half_partial) + logging.warning( + "`RuntimeOption.set_ipu_config` will be deprecated in v1.2.0, please use `RuntimeOption.paddle_infer_option.set_ipu_config()` instead." + ) + self._option.paddle_infer_option.set_ipu_config( + enable_fp16, replica_num, available_memory_proportion, + enable_half_partial) @property def poros_option(self): @@ -657,6 +659,11 @@ class RuntimeOption: """ return self._option.disable_profiling() + def set_external_raw_stream(self, cuda_stream): + """Set the external raw stream used by fastdeploy runtime. + """ + self._option.set_external_raw_stream(cuda_stream) + def __repr__(self): attrs = dir(self._option) message = "RuntimeOption(\n" From abae858f09b411af1ee840117531d1af08c1e237 Mon Sep 17 00:00:00 2001 From: chenjian Date: Tue, 14 Feb 2023 14:28:23 +0800 Subject: [PATCH 25/30] [C api] Add install command for c api and a demo in paddledection (#1289) * add c api demo and fix bugs * fix doc * add config.h.in * add config.h.in * add config.h.in --- CMakeLists.txt | 10 + FastDeploy.cmake.in | 3 + c_api/CMakeLists.txt | 1 + c_api/fastdeploy_capi/config.h | 22 ++ c_api/fastdeploy_capi/config.h.in | 22 ++ c_api/fastdeploy_capi/enum_variables.h | 71 ++++++ c_api/fastdeploy_capi/fd_common.h | 40 ---- c_api/fastdeploy_capi/fd_type.cc | 40 ++++ c_api/fastdeploy_capi/fd_type.h | 19 +- c_api/fastdeploy_capi/runtime_option.cc | 17 +- c_api/fastdeploy_capi/runtime_option.h | 17 +- c_api/fastdeploy_capi/vision.h | 26 +++ .../vision/classification/ppcls/model.cc | 7 +- .../vision/detection/ppdet/model.cc | 7 +- c_api/fastdeploy_capi/vision/result.cc | 6 +- c_api/fastdeploy_capi/vision/result.h | 16 -- c_api/fastdeploy_capi/vision/visualize.cc | 6 +- .../paddledetection/c/CMakeLists.txt | 13 ++ .../detection/paddledetection/c/README.md | 200 +++++++++++++++++ .../detection/paddledetection/c/README_CN.md | 204 ++++++++++++++++++ .../paddledetection/c/infer_ppyoloe.c | 124 +++++++++++ 21 files changed, 782 insertions(+), 89 deletions(-) create mode 100755 c_api/fastdeploy_capi/config.h create mode 100755 c_api/fastdeploy_capi/config.h.in create mode 100644 c_api/fastdeploy_capi/enum_variables.h create mode 100644 c_api/fastdeploy_capi/fd_type.cc create mode 100644 c_api/fastdeploy_capi/vision.h create mode 100644 examples/vision/detection/paddledetection/c/CMakeLists.txt create mode 100644 examples/vision/detection/paddledetection/c/README.md create mode 100644 examples/vision/detection/paddledetection/c/README_CN.md create mode 100644 examples/vision/detection/paddledetection/c/infer_ppyoloe.c diff --git a/CMakeLists.txt b/CMakeLists.txt index 51ba10c83..4f85653f2 100755 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -715,6 +715,16 @@ if(WITH_ASCEND) ) endif() +if(WITH_CAPI) + install( + DIRECTORY ${PROJECT_SOURCE_DIR}/c_api/fastdeploy_capi + DESTINATION ${CMAKE_INSTALL_PREFIX}/include + FILES_MATCHING + PATTERN "*.h" + PATTERN "*/types_internal.h" EXCLUDE + ) +endif() + include(${PROJECT_SOURCE_DIR}/cmake/config_cpack.cmake) ############################### Building: FastDeploy Python Wheel ############################# diff --git a/FastDeploy.cmake.in b/FastDeploy.cmake.in index d622660f4..c79001c28 100644 --- a/FastDeploy.cmake.in +++ b/FastDeploy.cmake.in @@ -33,6 +33,7 @@ set(ORT_DIRECTORY "@ORT_DIRECTORY@") set(OPENVINO_DIRECTORY "@OPENVINO_DIRECTORY@") set(RKNN2_TARGET_SOC "@RKNN2_TARGET_SOC@") set(WITH_KUNLUNXIN @WITH_KUNLUNXIN@) +set(WITH_CAPI @WITH_CAPI@) # Whether to use FastDeploy static lib. The default # value for this option is determined by the SDK # build-time options. @@ -357,6 +358,7 @@ message(STATUS " CMAKE_INSTALL_PREFIX : ${CMAKE_INSTALL_PREFIX}") message(STATUS " CMAKE_MODULE_PATH : ${CMAKE_MODULE_PATH}") message(STATUS "") message(STATUS " WITH_GPU : ${WITH_GPU}") +message(STATUS " WITH_CAPI : ${WITH_CAPI}") message(STATUS " ENABLE_ORT_BACKEND : ${ENABLE_ORT_BACKEND}") message(STATUS " ENABLE_RKNPU2_BACKEND : ${ENABLE_RKNPU2_BACKEND}") message(STATUS " ENABLE_SOPHGO_BACKEND : ${ENABLE_SOPHGO_BACKEND}") @@ -365,6 +367,7 @@ message(STATUS " ENABLE_POROS_BACKEND : ${ENABLE_POROS_BACKEND}") message(STATUS " ENABLE_OPENVINO_BACKEND : ${ENABLE_OPENVINO_BACKEND}") message(STATUS " ENABLE_TRT_BACKEND : ${ENABLE_TRT_BACKEND}") message(STATUS " ENABLE_LITE_BACKEND : ${ENABLE_LITE_BACKEND}") + if(ENABLE_PADDLE_BACKEND) message(STATUS " Paddle Inference version : ${PADDLEINFERENCE_VERSION}") endif() diff --git a/c_api/CMakeLists.txt b/c_api/CMakeLists.txt index 7c7a16626..4f3934165 100644 --- a/c_api/CMakeLists.txt +++ b/c_api/CMakeLists.txt @@ -19,6 +19,7 @@ if(NOT WITH_CAPI) return() endif() +configure_file(${PROJECT_SOURCE_DIR}/${CSRCS_DIR_NAME}/c_api/fastdeploy_capi/config.h.in ${PROJECT_SOURCE_DIR}/${CSRCS_DIR_NAME}/c_api/fastdeploy_capi/config.h) file(GLOB_RECURSE DEPLOY_CAPI_SRCS ${PROJECT_SOURCE_DIR}/${CSRCS_DIR_NAME}/c_api/fastdeploy_capi/*.cc) if(NOT ENABLE_VISION) file(GLOB_RECURSE DEPLOY_VISION_CAPI_SRCS ${PROJECT_SOURCE_DIR}/${CSRCS_DIR_NAME}/c_api/fastdeploy_capi/vision/*.cc) diff --git a/c_api/fastdeploy_capi/config.h b/c_api/fastdeploy_capi/config.h new file mode 100755 index 000000000..73de04c44 --- /dev/null +++ b/c_api/fastdeploy_capi/config.h @@ -0,0 +1,22 @@ +// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. +#pragma once + +#ifndef ENABLE_VISION +#define ENABLE_VISION +#endif + +#ifndef ENABLE_TEXT +/* #undef ENABLE_TEXT */ +#endif diff --git a/c_api/fastdeploy_capi/config.h.in b/c_api/fastdeploy_capi/config.h.in new file mode 100755 index 000000000..4b5b0137c --- /dev/null +++ b/c_api/fastdeploy_capi/config.h.in @@ -0,0 +1,22 @@ +// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. +#pragma once + +#ifndef ENABLE_VISION +#cmakedefine ENABLE_VISION +#endif + +#ifndef ENABLE_TEXT +#cmakedefine ENABLE_TEXT +#endif diff --git a/c_api/fastdeploy_capi/enum_variables.h b/c_api/fastdeploy_capi/enum_variables.h new file mode 100644 index 000000000..05a3d4c9f --- /dev/null +++ b/c_api/fastdeploy_capi/enum_variables.h @@ -0,0 +1,71 @@ +// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#pragma once + +#define FD_ENUM(type) \ + typedef int32_t type; \ + enum + +FD_ENUM(FD_C_ModelFormat){ + AUTOREC, ///< Auto recognize the model format by model file name + PADDLE, ///< Model with paddlepaddle format + ONNX, ///< Model with ONNX format + RKNN, ///< Model with RKNN format + TORCHSCRIPT, ///< Model with TorchScript format + SOPHGO, ///< Model with SOPHGO format +}; + +FD_ENUM(FD_C_rknpu2_CpuName){ + RK356X = 0, /* run on RK356X. */ + RK3588 = 1, /* default,run on RK3588. */ + UNDEFINED, +}; + +FD_ENUM(FD_C_rknpu2_CoreMask){ + RKNN_NPU_CORE_AUTO = 0, //< default, run on NPU core randomly. + RKNN_NPU_CORE_0 = 1, //< run on NPU core 0. + RKNN_NPU_CORE_1 = 2, //< run on NPU core 1. + RKNN_NPU_CORE_2 = 4, //< run on NPU core 2. + RKNN_NPU_CORE_0_1 = RKNN_NPU_CORE_0 | + RKNN_NPU_CORE_1, //< run on NPU core 1 and core 2. + RKNN_NPU_CORE_0_1_2 = RKNN_NPU_CORE_0_1 | + RKNN_NPU_CORE_2, //< run on NPU core 1 and core 2. + RKNN_NPU_CORE_UNDEFINED, +}; + +FD_ENUM(FD_C_LitePowerMode){ + LITE_POWER_HIGH = 0, ///< Use Lite Backend with high power mode + LITE_POWER_LOW = 1, ///< Use Lite Backend with low power mode + LITE_POWER_FULL = 2, ///< Use Lite Backend with full power mode + LITE_POWER_NO_BIND = 3, ///< Use Lite Backend with no bind power mode + LITE_POWER_RAND_HIGH = 4, ///< Use Lite Backend with rand high mode + LITE_POWER_RAND_LOW = 5 ///< Use Lite Backend with rand low power mode +}; + +FD_ENUM(FD_C_ResultType){ + UNKNOWN_RESULT, + CLASSIFY, + DETECTION, + SEGMENTATION, + OCR, + MOT, + FACE_DETECTION, + FACE_ALIGNMENT, + FACE_RECOGNITION, + MATTING, + MASK, + KEYPOINT_DETECTION, + HEADPOSE, +}; diff --git a/c_api/fastdeploy_capi/fd_common.h b/c_api/fastdeploy_capi/fd_common.h index 6374cf9b5..18bfb3c91 100644 --- a/c_api/fastdeploy_capi/fd_common.h +++ b/c_api/fastdeploy_capi/fd_common.h @@ -58,43 +58,3 @@ typedef int8_t FD_C_Bool; #define TRUE 1 #define FALSE 0 - -#define FD_ENUM(type) \ - typedef int32_t type; \ - enum - -FD_ENUM(FD_C_ModelFormat){ - AUTOREC, ///< Auto recognize the model format by model file name - PADDLE, ///< Model with paddlepaddle format - ONNX, ///< Model with ONNX format - RKNN, ///< Model with RKNN format - TORCHSCRIPT, ///< Model with TorchScript format - SOPHGO, ///< Model with SOPHGO format -}; - -FD_ENUM(FD_C_rknpu2_CpuName){ - RK356X = 0, /* run on RK356X. */ - RK3588 = 1, /* default,run on RK3588. */ - UNDEFINED, -}; - -FD_ENUM(FD_C_rknpu2_CoreMask){ - RKNN_NPU_CORE_AUTO = 0, //< default, run on NPU core randomly. - RKNN_NPU_CORE_0 = 1, //< run on NPU core 0. - RKNN_NPU_CORE_1 = 2, //< run on NPU core 1. - RKNN_NPU_CORE_2 = 4, //< run on NPU core 2. - RKNN_NPU_CORE_0_1 = RKNN_NPU_CORE_0 | - RKNN_NPU_CORE_1, //< run on NPU core 1 and core 2. - RKNN_NPU_CORE_0_1_2 = RKNN_NPU_CORE_0_1 | - RKNN_NPU_CORE_2, //< run on NPU core 1 and core 2. - RKNN_NPU_CORE_UNDEFINED, -}; - -FD_ENUM(FD_C_LitePowerMode){ - LITE_POWER_HIGH = 0, ///< Use Lite Backend with high power mode - LITE_POWER_LOW = 1, ///< Use Lite Backend with low power mode - LITE_POWER_FULL = 2, ///< Use Lite Backend with full power mode - LITE_POWER_NO_BIND = 3, ///< Use Lite Backend with no bind power mode - LITE_POWER_RAND_HIGH = 4, ///< Use Lite Backend with rand high mode - LITE_POWER_RAND_LOW = 5 ///< Use Lite Backend with rand low power mode -}; diff --git a/c_api/fastdeploy_capi/fd_type.cc b/c_api/fastdeploy_capi/fd_type.cc new file mode 100644 index 000000000..34390888a --- /dev/null +++ b/c_api/fastdeploy_capi/fd_type.cc @@ -0,0 +1,40 @@ +// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include "fastdeploy_capi/fd_type.h" + +#include + +#include "fastdeploy_capi/fd_common.h" + +#ifdef __cplusplus +extern "C" { +#endif + +FD_C_Mat FD_C_Imread(const char* imgpath) { + cv::Mat image = cv::imread(imgpath); + return new cv::Mat(image); +} + +FD_C_Bool FD_C_Imwrite(const char* savepath, FD_C_Mat img) { + cv::Mat cv_img = *(reinterpret_cast(img)); + bool result = cv::imwrite(savepath, cv_img); + return result; +} + +void FD_C_DestroyMat(FD_C_Mat mat) { delete reinterpret_cast(mat); } + +#ifdef __cplusplus +} +#endif diff --git a/c_api/fastdeploy_capi/fd_type.h b/c_api/fastdeploy_capi/fd_type.h index 75daf9db6..4a3d8cadd 100644 --- a/c_api/fastdeploy_capi/fd_type.h +++ b/c_api/fastdeploy_capi/fd_type.h @@ -17,7 +17,8 @@ #include #include -#include "fastdeploy_capi/fd_common.h" // NOLINT +#include "fastdeploy_capi/enum_variables.h" +#include "fastdeploy_capi/fd_common.h" typedef struct FD_C_OneDimArrayUint8 { size_t size; @@ -65,3 +66,19 @@ typedef struct FD_C_TwoDimArrayFloat { } FD_C_TwoDimArrayFloat; // std::vector> typedef void* FD_C_Mat; + +#ifdef __cplusplus +extern "C" { +#endif + +FASTDEPLOY_CAPI_EXPORT extern __fd_give FD_C_Mat +FD_C_Imread(const char* imgpath); + +FASTDEPLOY_CAPI_EXPORT extern FD_C_Bool FD_C_Imwrite(const char* savepath, + __fd_keep FD_C_Mat); + +FASTDEPLOY_CAPI_EXPORT extern void FD_C_DestroyMat(__fd_take FD_C_Mat mat); + +#ifdef __cplusplus +} +#endif diff --git a/c_api/fastdeploy_capi/runtime_option.cc b/c_api/fastdeploy_capi/runtime_option.cc index 3c9b4022d..4683d468d 100644 --- a/c_api/fastdeploy_capi/runtime_option.cc +++ b/c_api/fastdeploy_capi/runtime_option.cc @@ -17,7 +17,9 @@ #include "fastdeploy/utils/utils.h" #include "fastdeploy_capi/types_internal.h" +#ifdef __cplusplus extern "C" { +#endif FD_C_RuntimeOptionWrapper* FD_C_CreateRuntimeOptionWrapper() { FD_C_RuntimeOptionWrapper* fd_c_runtime_option_wrapper = @@ -28,7 +30,7 @@ FD_C_RuntimeOptionWrapper* FD_C_CreateRuntimeOptionWrapper() { return fd_c_runtime_option_wrapper; } -void FD_C_DestroyRuntimeOption( +void FD_C_DestroyRuntimeOptionWrapper( __fd_take FD_C_RuntimeOptionWrapper* fd_c_runtime_option_wrapper) { delete fd_c_runtime_option_wrapper; } @@ -404,15 +406,6 @@ void FD_C_RuntimeOptionWrapperUseIpu( batches_per_step); } -void FD_C_RuntimeOptionWrapperSetIpuConfig( - __fd_keep FD_C_RuntimeOptionWrapper* fd_c_runtime_option_wrapper, - FD_C_Bool enable_fp16, int replica_num, float available_memory_proportion, - FD_C_Bool enable_half_partial) { - auto& runtime_option = CHECK_AND_CONVERT_FD_TYPE(RuntimeOptionWrapper, - fd_c_runtime_option_wrapper); - runtime_option->SetIpuConfig(enable_fp16, replica_num, - available_memory_proportion, - enable_half_partial); +#ifdef __cplusplus } - -} // extern "C" +#endif diff --git a/c_api/fastdeploy_capi/runtime_option.h b/c_api/fastdeploy_capi/runtime_option.h index cfc087473..40d220bbe 100644 --- a/c_api/fastdeploy_capi/runtime_option.h +++ b/c_api/fastdeploy_capi/runtime_option.h @@ -14,9 +14,7 @@ #pragma once -#include - -#include "fastdeploy_capi/fd_common.h" +#include "fastdeploy_capi/fd_type.h" typedef struct FD_C_RuntimeOptionWrapper FD_C_RuntimeOptionWrapper; @@ -499,19 +497,6 @@ FASTDEPLOY_CAPI_EXPORT extern void FD_C_RuntimeOptionWrapperUseIpu( int device_num, int micro_batch_size, FD_C_Bool enable_pipelining, int batches_per_step); -/** \brief Set IPU config. - * - * \param[in] fd_c_runtime_option_wrapper pointer to FD_C_RuntimeOptionWrapper object - * \param[in] enable_fp16 enable fp16. - * \param[in] replica_num the number of graph replication. - * \param[in] available_memory_proportion the available memory proportion for matmul/conv. - * \param[in] enable_half_partial enable fp16 partial for matmul, only work with fp16. - */ -FASTDEPLOY_CAPI_EXPORT extern void FD_C_RuntimeOptionWrapperSetIpuConfig( - __fd_keep FD_C_RuntimeOptionWrapper* fd_c_runtime_option_wrapper, - FD_C_Bool enable_fp16, int replica_num, float available_memory_proportion, - FD_C_Bool enable_half_partial); - #ifdef __cplusplus } // extern "C" #endif diff --git a/c_api/fastdeploy_capi/vision.h b/c_api/fastdeploy_capi/vision.h new file mode 100644 index 000000000..a27a6c8dd --- /dev/null +++ b/c_api/fastdeploy_capi/vision.h @@ -0,0 +1,26 @@ +// Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. +#pragma once + +#include "fastdeploy_capi/config.h" + +#ifdef ENABLE_VISION +#include "fastdeploy_capi/vision/classification/ppcls/model.h" +#include "fastdeploy_capi/vision/detection/ppdet/model.h" +#include "fastdeploy_capi/vision/result.h" +#include "fastdeploy_capi/vision/visualize.h" +#endif + +#include "fastdeploy_capi/fd_type.h" +#include "fastdeploy_capi/runtime_option.h" diff --git a/c_api/fastdeploy_capi/vision/classification/ppcls/model.cc b/c_api/fastdeploy_capi/vision/classification/ppcls/model.cc index 3ed62f26a..ad0028612 100644 --- a/c_api/fastdeploy_capi/vision/classification/ppcls/model.cc +++ b/c_api/fastdeploy_capi/vision/classification/ppcls/model.cc @@ -16,7 +16,9 @@ #include "fastdeploy_capi/types_internal.h" +#ifdef __cplusplus extern "C" { +#endif FD_C_PaddleClasModelWrapper* FD_C_CreatePaddleClasModelWrapper( const char* model_file, const char* params_file, const char* config_file, @@ -50,4 +52,7 @@ FD_C_Bool FD_C_PaddleClasModelWrapperPredict( ClassifyResultWrapper, fd_c_classify_result_wrapper); return paddleclas_model->Predict(im, classify_result.get()); } -} \ No newline at end of file + +#ifdef __cplusplus +} +#endif \ No newline at end of file diff --git a/c_api/fastdeploy_capi/vision/detection/ppdet/model.cc b/c_api/fastdeploy_capi/vision/detection/ppdet/model.cc index 17a87ec8b..e7055185f 100644 --- a/c_api/fastdeploy_capi/vision/detection/ppdet/model.cc +++ b/c_api/fastdeploy_capi/vision/detection/ppdet/model.cc @@ -17,7 +17,9 @@ #include "fastdeploy_capi/types_internal.h" #include "fastdeploy_capi/vision/visualize.h" +#ifdef __cplusplus extern "C" { +#endif FD_C_PPYOLOEWrapper* FD_C_CreatesPPYOLOEWrapper( const char* model_file, const char* params_file, const char* config_file, @@ -50,4 +52,7 @@ FD_C_Bool FD_C_PPYOLOEWrapperPredict( DetectionResultWrapper, fd_c_detection_result_wrapper); return ppyoloe_model->Predict(im, detection_result.get()); } -} \ No newline at end of file + +#ifdef __cplusplus +} +#endif \ No newline at end of file diff --git a/c_api/fastdeploy_capi/vision/result.cc b/c_api/fastdeploy_capi/vision/result.cc index abf52ba69..207ef3950 100644 --- a/c_api/fastdeploy_capi/vision/result.cc +++ b/c_api/fastdeploy_capi/vision/result.cc @@ -17,7 +17,9 @@ #include "fastdeploy/utils/utils.h" #include "fastdeploy_capi/types_internal.h" +#ifdef __cplusplus extern "C" { +#endif // Classification Results @@ -235,4 +237,6 @@ FD_C_DetectionResultWrapper* FD_C_CreateDetectionResultWrapperFromData( return fd_c_detection_result_wrapper; } -} \ No newline at end of file +#ifdef __cplusplus +} +#endif \ No newline at end of file diff --git a/c_api/fastdeploy_capi/vision/result.h b/c_api/fastdeploy_capi/vision/result.h index 9d32052d9..247ec50f6 100644 --- a/c_api/fastdeploy_capi/vision/result.h +++ b/c_api/fastdeploy_capi/vision/result.h @@ -24,22 +24,6 @@ typedef struct FD_C_DetectionResultWrapper FD_C_DetectionResultWrapper; extern "C" { #endif -FD_ENUM(FD_C_ResultType){ - UNKNOWN_RESULT, - CLASSIFY, - DETECTION, - SEGMENTATION, - OCR, - MOT, - FACE_DETECTION, - FACE_ALIGNMENT, - FACE_RECOGNITION, - MATTING, - MASK, - KEYPOINT_DETECTION, - HEADPOSE, -}; - typedef struct FD_C_ClassifyResult { FD_C_OneDimArrayInt32 label_ids; FD_C_OneDimArrayFloat scores; diff --git a/c_api/fastdeploy_capi/vision/visualize.cc b/c_api/fastdeploy_capi/vision/visualize.cc index 9132fe606..6045270bd 100644 --- a/c_api/fastdeploy_capi/vision/visualize.cc +++ b/c_api/fastdeploy_capi/vision/visualize.cc @@ -17,7 +17,9 @@ #include "fastdeploy/vision/visualize/visualize.h" #include "fastdeploy_capi/types_internal.h" +#ifdef __cplusplus extern "C" { +#endif FD_C_Mat FD_C_VisDetection(FD_C_Mat im, FD_C_DetectionResult* fd_c_detection_result, @@ -32,4 +34,6 @@ FD_C_Mat FD_C_VisDetection(FD_C_Mat im, line_size, font_size); return new cv::Mat(result); } -} \ No newline at end of file +#ifdef __cplusplus +} +#endif \ No newline at end of file diff --git a/examples/vision/detection/paddledetection/c/CMakeLists.txt b/examples/vision/detection/paddledetection/c/CMakeLists.txt new file mode 100644 index 000000000..12a8d9a2c --- /dev/null +++ b/examples/vision/detection/paddledetection/c/CMakeLists.txt @@ -0,0 +1,13 @@ +PROJECT(infer_demo C) +CMAKE_MINIMUM_REQUIRED (VERSION 3.10) + +# 指定下载解压后的fastdeploy库路径 +option(FASTDEPLOY_INSTALL_DIR "Path of downloaded fastdeploy sdk.") + +include(${FASTDEPLOY_INSTALL_DIR}/FastDeploy.cmake) + +# 添加FastDeploy依赖头文件 +include_directories(${FASTDEPLOY_INCS}) + +add_executable(infer_ppyoloe_demo ${PROJECT_SOURCE_DIR}/infer_ppyoloe.c) +target_link_libraries(infer_ppyoloe_demo ${FASTDEPLOY_LIBS}) diff --git a/examples/vision/detection/paddledetection/c/README.md b/examples/vision/detection/paddledetection/c/README.md new file mode 100644 index 000000000..79a33e51a --- /dev/null +++ b/examples/vision/detection/paddledetection/c/README.md @@ -0,0 +1,200 @@ +English | [简体中文](README_CN.md) +# PaddleDetection C Deployment Example + +This directory provides examples that `infer_xxx.c` fast finishes the deployment of PaddleDetection models, including PPYOLOE on CPU/GPU. + +Before deployment, two steps require confirmation + +- 1. Software and hardware should meet the requirements. Please refer to [FastDeploy Environment Requirements](../../../../../docs/en/build_and_install/download_prebuilt_libraries.md) +- 2. Download the precompiled deployment library and samples code according to your development environment. Refer to [FastDeploy Precompiled Library](../../../../../docs/en/build_and_install/download_prebuilt_libraries.md) + +Taking inference on Linux as an example, the compilation test can be completed by executing the following command in this directory. FastDeploy version 1.0.4 or above (x.x.x>=1.0.4) is required to support this model. + +```bash +ppyoloe is taken as an example for inference deployment + +mkdir build +cd build +# Download the FastDeploy precompiled library. Users can choose your appropriate version in the `FastDeploy Precompiled Library` mentioned above +wget https://bj.bcebos.com/fastdeploy/release/cpp/fastdeploy-linux-x64-x.x.x.tgz +tar xvf fastdeploy-linux-x64-x.x.x.tgz +cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/fastdeploy-linux-x64-x.x.x +make -j + +# Download the PPYOLOE model file and test images +wget https://bj.bcebos.com/paddlehub/fastdeploy/ppyoloe_crn_l_300e_coco.tgz +wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg +tar xvf ppyoloe_crn_l_300e_coco.tgz + + +# CPU inference +./infer_ppyoloe_demo ./ppyoloe_crn_l_300e_coco 000000014439.jpg 0 +# GPU inference +./infer_ppyoloe_demo ./ppyoloe_crn_l_300e_coco 000000014439.jpg 1 +``` + +The above command works for Linux or MacOS. For SDK use-pattern in Windows, refer to: +- [How to use FastDeploy C++ SDK in Windows](../../../../../docs/en/faq/use_sdk_on_windows.md) + +## PaddleDetection C Interface + +### RuntimeOption + +```c +FD_C_RuntimeOptionWrapper* FD_C_CreateRuntimeOptionWrapper() +``` + +> Create a RuntimeOption object, and return a pointer to manipulate it. +> +> **Return** +> * **fd_c_runtime_option_wrapper**(FD_C_RuntimeOptionWrapper*): Pointer to manipulate RuntimeOption object. + + +```c +void FD_C_RuntimeOptionWrapperUseCpu( + FD_C_RuntimeOptionWrapper* fd_c_runtime_option_wrapper) +``` + +> Enable Cpu inference. +> +> **Params** +> +> * **fd_c_runtime_option_wrapper**(FD_C_RuntimeOptionWrapper*): Pointer to manipulate RuntimeOption object. + +```c +void FD_C_RuntimeOptionWrapperUseGpu( + FD_C_RuntimeOptionWrapper* fd_c_runtime_option_wrapper, + int gpu_id) +``` +> Enable Gpu inference. +> +> **Params** +> +> * **fd_c_runtime_option_wrapper**(FD_C_RuntimeOptionWrapper*): Pointer to manipulate RuntimeOption object. +> * **gpu_id**(int): gpu id + + +### Model + +```c + +FD_C_PPYOLOEWrapper* FD_C_CreatesPPYOLOEWrapper( + const char* model_file, const char* params_file, const char* config_file, + FD_C_RuntimeOptionWrapper* runtime_option, + const FD_C_ModelFormat model_format) + +``` + +> Create a PPYOLOE model object, and return a pointer to manipulate it. +> +> **Params** +> +> * **model_file**(const char*): Model file path +> * **params_file**(const char*): Parameter file path +> * **config_file**(const char*): Configuration file path, which is the deployment yaml file exported by PaddleDetection +> * **runtime_option**(FD_C_RuntimeOptionWrapper*): Backend inference configuration. None by default, which is the default configuration +> * **model_format**(FD_C_ModelFormat): Model format. Paddle format by default +> +> **Return** +> * **fd_c_ppyoloe_wrapper**(FD_C_PPYOLOEWrapper*): Pointer to manipulate PPYOLOE object. + + +#### Read and write image + +```c +FD_C_Mat FD_C_Imread(const char* imgpath) +``` + +> Read an image, and return a pointer to cv::Mat. +> +> **Params** +> +> * **imgpath**(const char*): image path +> +> **Return** +> +> * **imgmat**(FD_C_Mat): pointer to cv::Mat object which holds the image. + + +```c +FD_C_Bool FD_C_Imwrite(const char* savepath, FD_C_Mat img); +``` + +> Write image to a file. +> +> **Params** +> +> * **savepath**(const char*): save path +> * **img**(FD_C_Mat): pointer to cv::Mat object +> +> **Return** +> +> * **result**(FD_C_Bool): bool to indicate success or failure + + +#### Prediction + +```c +FD_C_Bool FD_C_PPYOLOEWrapperPredict( + __fd_take FD_C_PPYOLOEWrapper* fd_c_ppyoloe_wrapper, FD_C_Mat img, + FD_C_DetectionResultWrapper* fd_c_detection_result_wrapper) +``` +> +> Predict an image, and generate detection result. +> +> **Params** +> * **fd_c_ppyoloe_wrapper**(FD_C_PPYOLOEWrapper*): pointer to manipulate PPYOLOE object +> * **img**（FD_C_Mat）: pointer to cv::Mat object, which can be obained by FD_C_Imread interface +> * **result**（FD_C_DetectionResultWrapper*): Detection result, including detection box and confidence of each box. Refer to [Vision Model Prediction Result](../../../../../docs/api/vision_results/) for DetectionResult + + +#### Result + +```c +FD_C_DetectionResultWrapper* FD_C_CreateDetectionResultWrapper(); +``` +> +> Create a DetectionResult object to keep the detection result，return a pointer to manipulate it. +> +> **Return** +> * **fd_c_detection_result_wrapper**(FD_C_DetectionResultWrapper*): pointer to manipulate DetectionResult object + + + +```c +FD_C_DetectionResult* FD_C_DetectionResultWrapperGetData( + FD_C_DetectionResultWrapper* fd_c_detection_result_wrapper) +``` +> +> Get the C DetectionResult structure from FD_C_DetectionResultWrapper, which can access the fileds directly. +> +> **Params** +> * **fd_c_detection_result_wrapper**(FD_C_DetectionResultWrapper*): pointer to manipulate DetectionResult object +> +> **Return** +> * **fd_c_detection_result**(FD_C_DetectionResult*): pointer to C DetectionResult structure + + + +```c +FD_C_Mat FD_C_VisDetection(FD_C_Mat im, FD_C_DetectionResult* fd_detection_result, + float score_threshold, int line_size, float font_size); +``` +> +> Visualize detection results and return visualization image. +> +> **Params** +> * **im**(FD_C_Mat): pointer to input image +> * **fd_detection_result**(FD_C_DetectionResult*): pointer to C DetectionResult structure +> * **score_threshold**(float): score threshold +> * **line_size**(int): line size +> * **font_size**(float): font size +> +> **Return** +> * **vis_im**(FD_C_Mat): pointer to visualization image. + + +- [Model Description](../../) +- [Python Deployment](../python) +- [Vision Model prediction results](../../../../../docs/api/vision_results/) +- [How to switch the model inference backend engine](../../../../../docs/en/faq/how_to_change_backend.md) diff --git a/examples/vision/detection/paddledetection/c/README_CN.md b/examples/vision/detection/paddledetection/c/README_CN.md new file mode 100644 index 000000000..22cbd311c --- /dev/null +++ b/examples/vision/detection/paddledetection/c/README_CN.md @@ -0,0 +1,204 @@ +[English](README.md) | 简体中文 +# PaddleDetection C 部署示例 + +本目录下提供`infer_xxx.c`来调用C API快速完成PaddleDetection模型PPYOLOE在CPU/GPU上部署的示例。 + +在部署前，需确认以下两个步骤 + +- 1. 软硬件环境满足要求，参考[FastDeploy环境要求](../../../../../docs/cn/build_and_install/download_prebuilt_libraries.md) +- 2. 根据开发环境，下载预编译部署库和samples代码，参考[FastDeploy预编译库](../../../../../docs/cn/build_and_install/download_prebuilt_libraries.md) + +以Linux上推理为例，在本目录执行如下命令即可完成编译测试，支持此模型需保证FastDeploy版本1.0.4以上(x.x.x>=1.0.4) + +```bash +以ppyoloe为例进行推理部署 + +mkdir build +cd build +# 下载FastDeploy预编译库，用户可在上文提到的`FastDeploy预编译库`中自行选择合适的版本使用 +wget https://bj.bcebos.com/fastdeploy/release/cpp/fastdeploy-linux-x64-x.x.x.tgz +tar xvf fastdeploy-linux-x64-x.x.x.tgz +cmake .. -DFASTDEPLOY_INSTALL_DIR=${PWD}/fastdeploy-linux-x64-x.x.x +make -j + +# 下载PPYOLOE模型文件和测试图片 +wget https://bj.bcebos.com/paddlehub/fastdeploy/ppyoloe_crn_l_300e_coco.tgz +wget https://gitee.com/paddlepaddle/PaddleDetection/raw/release/2.4/demo/000000014439.jpg +tar xvf ppyoloe_crn_l_300e_coco.tgz + + +# CPU推理 +./infer_ppyoloe_demo ./ppyoloe_crn_l_300e_coco 000000014439.jpg 0 +# GPU推理 +./infer_ppyoloe_demo ./ppyoloe_crn_l_300e_coco 000000014439.jpg 1 +``` + +以上命令只适用于Linux或MacOS, Windows下SDK的使用方式请参考: +- [如何在Windows中使用FastDeploy C++ SDK](../../../../../docs/cn/faq/use_sdk_on_windows.md) + +如果用户使用华为昇腾NPU部署, 请参考以下方式在部署前初始化部署环境: +- [如何使用华为昇腾NPU部署](../../../../../docs/cn/faq/use_sdk_on_ascend.md) + +## PaddleDetection C API接口 + +### 配置 + +```c +FD_C_RuntimeOptionWrapper* FD_C_CreateRuntimeOptionWrapper() +``` + +> 创建一个RuntimeOption的配置对象，并且返回操作它的指针。 +> +> **返回** +> +> * **fd_c_runtime_option_wrapper**(FD_C_RuntimeOptionWrapper*): 指向RuntimeOption对象的指针 + + +```c +void FD_C_RuntimeOptionWrapperUseCpu( + FD_C_RuntimeOptionWrapper* fd_c_runtime_option_wrapper) +``` + +> 开启CPU推理 +> +> **参数** +> +> * **fd_c_runtime_option_wrapper**(FD_C_RuntimeOptionWrapper*): 指向RuntimeOption对象的指针 + +```c +void FD_C_RuntimeOptionWrapperUseGpu( + FD_C_RuntimeOptionWrapper* fd_c_runtime_option_wrapper, + int gpu_id) +``` +> 开启GPU推理 +> +> **参数** +> +> * **fd_c_runtime_option_wrapper**(FD_C_RuntimeOptionWrapper*): 指向RuntimeOption对象的指针 +> * **gpu_id**(int): 显卡号 + + +### 模型 + +```c + +FD_C_PPYOLOEWrapper* FD_C_CreatesPPYOLOEWrapper( + const char* model_file, const char* params_file, const char* config_file, + FD_C_RuntimeOptionWrapper* runtime_option, + const FD_C_ModelFormat model_format) + +``` + +> 创建一个PPYOLOE的模型，并且返回操作它的指针。 +> +> **参数** +> +> * **model_file**(const char*): 模型文件路径 +> * **params_file**(const char*): 参数文件路径 +> * **config_file**(const char*): 配置文件路径，即PaddleDetection导出的部署yaml文件 +> * **runtime_option**(FD_C_RuntimeOptionWrapper*): 指向RuntimeOption的指针，表示后端推理配置 +> * **model_format**(FD_C_ModelFormat): 模型格式 +> +> **返回** +> * **fd_c_ppyoloe_wrapper**(FD_C_PPYOLOEWrapper*): 指向PPYOLOE模型对象的指针 + + +#### 读写图像 + +```c +FD_C_Mat FD_C_Imread(const char* imgpath) +``` + +> 读取一个图像，并且返回cv::Mat的指针。 +> +> **参数** +> +> * **imgpath**(const char*): 图像文件路径 +> +> **返回** +> +> * **imgmat**(FD_C_Mat): 指向图像数据cv::Mat的指针。 + + +```c +FD_C_Bool FD_C_Imwrite(const char* savepath, FD_C_Mat img); +``` + +> 将图像写入文件中。 +> +> **参数** +> +> * **savepath**(const char*): 保存图像的路径 +> * **img**(FD_C_Mat): 指向图像数据的指针 +> +> **返回** +> +> * **result**(FD_C_Bool): 表示操作是否成功 + + +#### Predict函数 + +```c +FD_C_Bool FD_C_PPYOLOEWrapperPredict( + __fd_take FD_C_PPYOLOEWrapper* fd_c_ppyoloe_wrapper, FD_C_Mat img, + FD_C_DetectionResultWrapper* fd_c_detection_result_wrapper) +``` +> +> 模型预测接口，输入图像直接并生成检测结果。 +> +> **参数** +> * **fd_c_ppyoloe_wrapper**(FD_C_PPYOLOEWrapper*): 指向PPYOLOE模型的指针 +> * **img**（FD_C_Mat）: 输入图像的指针，指向cv::Mat对象，可以调用FD_C_Imread读取图像获取 +> * **result**（FD_C_DetectionResultWrapper*): 指向检测结果的指针，检测结果包括检测框，各个框的置信度, DetectionResult说明参考[视觉模型预测结果](../../../../../docs/api/vision_results/) + + +#### Predict结果 + +```c +FD_C_DetectionResultWrapper* FD_C_CreateDetectionResultWrapper(); +``` +> +> 创建一个DetectionResult对象，用来保存推理的结果，并返回所创建的DetectionResult对象的指针。 +> +> **返回** +> * **fd_c_detection_result_wrapper**(FD_C_DetectionResultWrapper*): 指向DetectionResult对象的指针 + + + +```c +FD_C_DetectionResult* FD_C_DetectionResultWrapperGetData( + FD_C_DetectionResultWrapper* fd_c_detection_result_wrapper) +``` +> +> 从DetectionResult对象中提取纯C结构的DetectionResult结果，并返回结构指针，通过该指针可直接返回结构中的字段。 +> +> **参数** +> * **fd_c_detection_result_wrapper**(FD_C_DetectionResultWrapper*): 指向DetectionResult对象的指针 +> +> **返回** +> * **fd_c_detection_result**(FD_C_DetectionResult*): 指向纯C结构的DetectionResult的指针 + + + +```c +FD_C_Mat FD_C_VisDetection(FD_C_Mat im, FD_C_DetectionResult* fd_detection_result, + float score_threshold, int line_size, float font_size); +``` +> +> 对检测结果进行可视化，返回可视化的图像。 +> +> **参数** +> * **im**(FD_C_Mat): 指向输入图像的指针 +> * **fd_detection_result**(FD_C_DetectionResult*): 指向纯C结构DetectionResult的指针 +> * **score_threshold**(float): 检测阈值 +> * **line_size**(int): 检测框线大小 +> * **font_size**(float): 检测框字体大小 +> +> **返回** +> * **vis_im**(FD_C_Mat): 指向可视化图像的指针 + + +- [模型介绍](../../) +- [Python部署](../python) +- [视觉模型预测结果](../../../../../docs/api/vision_results/) +- [如何切换模型推理后端引擎](../../../../../docs/cn/faq/how_to_change_backend.md) diff --git a/examples/vision/detection/paddledetection/c/infer_ppyoloe.c b/examples/vision/detection/paddledetection/c/infer_ppyoloe.c new file mode 100644 index 000000000..6b9cc4369 --- /dev/null +++ b/examples/vision/detection/paddledetection/c/infer_ppyoloe.c @@ -0,0 +1,124 @@ +// Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved. +// +// Licensed under the Apache License, Version 2.0 (the "License"); +// you may not use this file except in compliance with the License. +// You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +#include +#include + +#include "fastdeploy_capi/vision.h" + +#ifdef WIN32 +const char sep = '\\'; +#else +const char sep = '/'; +#endif + +void CpuInfer(const char* model_dir, const char* image_file) { + char model_file[100]; + char params_file[100]; + char config_file[100]; + int max_size = 99; + snprintf(model_file, max_size, "%s%c%s", model_dir, sep, "model.pdmodel"); + snprintf(params_file, max_size, "%s%c%s", model_dir, sep, "model.pdiparams"); + snprintf(config_file, max_size, "%s%c%s", model_dir, sep, "infer_cfg.yml"); + + FD_C_RuntimeOptionWrapper* option = FD_C_CreateRuntimeOptionWrapper(); + FD_C_RuntimeOptionWrapperUseCpu(option); + + FD_C_PPYOLOEWrapper* model = FD_C_CreatesPPYOLOEWrapper( + model_file, params_file, config_file, option, PADDLE); + + FD_C_Mat im = FD_C_Imread(image_file); + + FD_C_DetectionResultWrapper* result_wrapper = + FD_C_CreateDetectionResultWrapper(); + + if (!FD_C_PPYOLOEWrapperPredict(model, im, result_wrapper)) { + printf("Failed to predict.\n"); + return; + } + + FD_C_DetectionResult* result = + FD_C_DetectionResultWrapperGetData(result_wrapper); + FD_C_Mat vis_im = FD_C_VisDetection(im, result, 0.5, 1, 0.5); + + FD_C_Imwrite("vis_result.jpg", vis_im); + printf("Visualized result saved in ./vis_result.jpg\n"); + + FD_C_DestroyRuntimeOptionWrapper(option); + FD_C_DestroyPPYOLOEWrapper(model); + FD_C_DestroyDetectionResultWrapper(result_wrapper); + FD_C_DestroyDetectionResult(result); + FD_C_DestroyMat(im); + FD_C_DestroyMat(vis_im); +} + +void GpuInfer(const char* model_dir, const char* image_file) { + char model_file[100]; + char params_file[100]; + char config_file[100]; + int max_size = 99; + snprintf(model_file, max_size, "%s%c%s", model_dir, sep, "model.pdmodel"); + snprintf(params_file, max_size, "%s%c%s", model_dir, sep, "model.pdiparams"); + snprintf(config_file, max_size, "%s%c%s", model_dir, sep, "infer_cfg.yml"); + + FD_C_RuntimeOptionWrapper* option = FD_C_CreateRuntimeOptionWrapper(); + FD_C_RuntimeOptionWrapperUseGpu(option, 0); + + FD_C_PPYOLOEWrapper* model = FD_C_CreatesPPYOLOEWrapper( + model_file, params_file, config_file, option, PADDLE); + + FD_C_Mat im = FD_C_Imread(image_file); + + FD_C_DetectionResultWrapper* result_wrapper = + FD_C_CreateDetectionResultWrapper(); + + if (!FD_C_PPYOLOEWrapperPredict(model, im, result_wrapper)) { + printf("Failed to predict.\n"); + return; + } + + FD_C_DetectionResult* result = + FD_C_DetectionResultWrapperGetData(result_wrapper); + FD_C_Mat vis_im = FD_C_VisDetection(im, result, 0.5, 1, 0.5); + + FD_C_Imwrite("vis_result.jpg", vis_im); + printf("Visualized result saved in ./vis_result.jpg\n"); + + FD_C_DestroyRuntimeOptionWrapper(option); + FD_C_DestroyPPYOLOEWrapper(model); + FD_C_DestroyDetectionResultWrapper(result_wrapper); + FD_C_DestroyDetectionResult(result); + FD_C_DestroyMat(im); + FD_C_DestroyMat(vis_im); +} + +int main(int argc, char* argv[]) { + if (argc < 4) { + printf( + "Usage: infer_demo path/to/model_dir path/to/image run_option, " + "e.g ./infer_model ./ppyoloe_model_dir ./test.jpeg 0" + "\n"); + printf( + "The data type of run_option is int, 0: run with cpu; 1: run with gpu" + "\n"); + return -1; + } + + if (atoi(argv[3]) == 0) { + CpuInfer(argv[1], argv[2]); + } else if (atoi(argv[3]) == 1) { + GpuInfer(argv[1], argv[2]); + } + return 0; +} From 2dfda1db85fdf9adf131f7fe82af4ac9a76fd9e7 Mon Sep 17 00:00:00 2001 From: WJJ1995 Date: Tue, 14 Feb 2023 15:00:05 +0800 Subject: [PATCH 26/30] [Benchmark] Remove pmap and use mem api (#1309) * add GPL lisence * add GPL-3.0 lisence * add GPL-3.0 lisence * add GPL-3.0 lisence * support yolov8 * add pybind for yolov8 * add yolov8 readme * add cpp benchmark * add cpu and gpu mem * public part split * add runtime mode * fixed bugs * add cpu_thread_nums * deal with comments * deal with comments * deal with comments * rm useless code * add FASTDEPLOY_DECL * add FASTDEPLOY_DECL * fixed for windows * mv rss to pss * mv rss to pss * Update utils.cc * use thread to collect mem * Add ResourceUsageMonitor * rm useless code * fixed bug * fixed typo * update ResourceUsageMonitor * fixed bug * fixed bug * add note for ResourceUsageMonitor * deal with comments * add macros * deal with comments * deal with comments * deal with comments * re-lint * rm pmap and use mem api * rm pmap and use mem api * add mem api * Add PrintBenchmarkInfo func * Add PrintBenchmarkInfo func * Add PrintBenchmarkInfo func * deal with comments --------- Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> --- benchmark/cpp/benchmark_ppyolov8.cc | 9 ++--- benchmark/cpp/benchmark_yolov5.cc | 3 +- benchmark/cpp/flags.h | 38 ++++++++++++++++++++ benchmark/cpp/macros.h | 5 ++- fastdeploy/benchmark/utils.cc | 56 +++++++++++------------------ fastdeploy/benchmark/utils.h | 14 +++++--- 6 files changed, 74 insertions(+), 51 deletions(-) mode change 100644 => 100755 benchmark/cpp/benchmark_ppyolov8.cc mode change 100644 => 100755 benchmark/cpp/benchmark_yolov5.cc diff --git a/benchmark/cpp/benchmark_ppyolov8.cc b/benchmark/cpp/benchmark_ppyolov8.cc old mode 100644 new mode 100755 index 545474635..5541696d5 --- a/benchmark/cpp/benchmark_ppyolov8.cc +++ b/benchmark/cpp/benchmark_ppyolov8.cc @@ -12,16 +12,10 @@ // See the License for the specific language governing permissions and // limitations under the License. -#include "flags.h" #include "macros.h" +#include "flags.h" #include "option.h" -#ifdef WIN32 -const char sep = '\\'; -#else -const char sep = '/'; -#endif - int main(int argc, char* argv[]) { google::ParseCommandLineFlags(&argc, &argv, true); auto im = cv::imread(FLAGS_image); @@ -31,6 +25,7 @@ int main(int argc, char* argv[]) { PrintUsage(); return false; } + PrintBenchmarkInfo(); auto model_file = FLAGS_model + sep + "model.pdmodel"; auto params_file = FLAGS_model + sep + "model.pdiparams"; auto config_file = FLAGS_model + sep + "infer_cfg.yml"; diff --git a/benchmark/cpp/benchmark_yolov5.cc b/benchmark/cpp/benchmark_yolov5.cc old mode 100644 new mode 100755 index 5b2cab855..6ab3c5990 --- a/benchmark/cpp/benchmark_yolov5.cc +++ b/benchmark/cpp/benchmark_yolov5.cc @@ -12,8 +12,8 @@ // See the License for the specific language governing permissions and // limitations under the License. -#include "flags.h" #include "macros.h" +#include "flags.h" #include "option.h" int main(int argc, char* argv[]) { @@ -25,6 +25,7 @@ int main(int argc, char* argv[]) { PrintUsage(); return false; } + PrintBenchmarkInfo(); auto model_yolov5 = fastdeploy::vision::detection::YOLOv5(FLAGS_model, "", option); fastdeploy::vision::DetectionResult res; diff --git a/benchmark/cpp/flags.h b/benchmark/cpp/flags.h index 7f8c3a29f..fd20e685c 100755 --- a/benchmark/cpp/flags.h +++ b/benchmark/cpp/flags.h @@ -16,6 +16,12 @@ #include "gflags/gflags.h" +#ifdef WIN32 +const char sep = '\\'; +#else +const char sep = '/'; +#endif + DEFINE_string(model, "", "Directory of the inference model."); DEFINE_string(image, "", "Path of the image file."); DEFINE_string(device, "cpu", @@ -48,3 +54,35 @@ void PrintUsage() { std::cout << "Default value of backend: default" << std::endl; std::cout << "Default value of use_fp16: false" << std::endl; } + +void PrintBenchmarkInfo() { + // Get model name + std::vector model_names; + fastdeploy::benchmark::Split(FLAGS_model, model_names, sep); + // Save benchmark info + std::stringstream ss; + ss.precision(3); + ss << "\n======= Model Info =======\n"; + ss << "model_name: " << model_names[model_names.size() - 1] << std::endl; + ss << "profile_mode: " << FLAGS_profile_mode << std::endl; + if (FLAGS_profile_mode == "runtime") { + ss << "include_h2d_d2h: " << FLAGS_include_h2d_d2h << std::endl; + } + ss << "\n======= Backend Info =======\n"; + ss << "warmup: " << FLAGS_warmup << std::endl; + ss << "repeats: " << FLAGS_repeat << std::endl; + ss << "device: " << FLAGS_device << std::endl; + if (FLAGS_device == "gpu") { + ss << "device_id: " << FLAGS_device_id << std::endl; + } + ss << "backend: " << FLAGS_backend << std::endl; + ss << "cpu_thread_nums: " << FLAGS_cpu_thread_nums << std::endl; + ss << "use_fp16: " << FLAGS_use_fp16 << std::endl; + ss << "collect_memory_info: " << FLAGS_collect_memory_info << std::endl; + if (FLAGS_collect_memory_info) { + ss << "sampling_interval: " << std::to_string(FLAGS_sampling_interval) + << "ms" << std::endl; + } + std::cout << ss.str() << std::endl; + return; +} diff --git a/benchmark/cpp/macros.h b/benchmark/cpp/macros.h index bebd26e0d..77df0c657 100755 --- a/benchmark/cpp/macros.h +++ b/benchmark/cpp/macros.h @@ -18,7 +18,6 @@ #define BENCHMARK_MODEL(MODEL_NAME, BENCHMARK_FUNC) \ { \ - std::cout << "====" << #MODEL_NAME << "====" << std::endl; \ if (!MODEL_NAME.Initialized()) { \ std::cerr << "Failed to initialize." << std::endl; \ return 0; \ @@ -62,8 +61,8 @@ float __cpu_mem__ = __resource_moniter__.GetMaxCpuMem(); \ float __gpu_mem__ = __resource_moniter__.GetMaxGpuMem(); \ float __gpu_util__ = __resource_moniter__.GetMaxGpuUtil(); \ - std::cout << "cpu_pss_mb: " << __cpu_mem__ << "MB." << std::endl; \ - std::cout << "gpu_pss_mb: " << __gpu_mem__ << "MB." << std::endl; \ + std::cout << "cpu_rss_mb: " << __cpu_mem__ << "MB." << std::endl; \ + std::cout << "gpu_rss_mb: " << __gpu_mem__ << "MB." << std::endl; \ std::cout << "gpu_util: " << __gpu_util__ << std::endl; \ __resource_moniter__.Stop(); \ } \ diff --git a/fastdeploy/benchmark/utils.cc b/fastdeploy/benchmark/utils.cc index a33db1dc2..8857f10c4 100755 --- a/fastdeploy/benchmark/utils.cc +++ b/fastdeploy/benchmark/utils.cc @@ -13,8 +13,8 @@ // limitations under the License. #include -#if defined(__linux__) || defined(__ANDROID__) -#include +#ifdef __linux__ +#include #endif #include @@ -23,8 +23,7 @@ namespace fastdeploy { namespace benchmark { -// Remove the ch characters at both ends of str -static std::string strip(const std::string& str, char ch = ' ') { +std::string Strip(const std::string& str, char ch) { int i = 0; while (str[i] == ch) { i++; @@ -36,9 +35,8 @@ static std::string strip(const std::string& str, char ch = ' ') { return str.substr(i, j + 1 - i); } -// Split string -static void split(const std::string& s, std::vector& tokens, - char delim = ' ') { +void Split(const std::string& s, std::vector& tokens, + char delim) { tokens.clear(); size_t lastPos = s.find_first_not_of(delim, 0); size_t pos = s.find(delim, lastPos); @@ -54,7 +52,7 @@ ResourceUsageMonitor::ResourceUsageMonitor(int sampling_interval_ms, int gpu_id) : is_supported_(false), sampling_interval_(sampling_interval_ms), gpu_id_(gpu_id) { -#if defined(__linux__) || defined(__ANDROID__) +#ifdef __linux__ is_supported_ = true; #else is_supported_ = false; @@ -67,7 +65,9 @@ ResourceUsageMonitor::ResourceUsageMonitor(int sampling_interval_ms, int gpu_id) } void ResourceUsageMonitor::Start() { - if (!is_supported_) return; + if (!is_supported_) { + return; + } if (check_memory_thd_ != nullptr) { FDINFO << "Memory monitoring has already started!" << std::endl; return; @@ -77,20 +77,24 @@ void ResourceUsageMonitor::Start() { check_memory_thd_.reset(new std::thread(([this]() { // Note we retrieve the memory usage at the very beginning of the thread. while (true) { - std::string cpu_mem_info = GetCurrentCpuMemoryInfo(); - // get max_cpu_mem - std::vector cpu_tokens; - split(cpu_mem_info, cpu_tokens, ' '); - max_cpu_mem_ = std::max(max_cpu_mem_, stof(cpu_tokens[3]) / 1024); +#ifdef __linux__ + rusage res; + if (getrusage(RUSAGE_SELF, &res) == 0) { + max_cpu_mem_ = + std::max(max_cpu_mem_, static_cast(res.ru_maxrss / 1024.0)); + } +#endif #if defined(WITH_GPU) std::string gpu_mem_info = GetCurrentGpuMemoryInfo(gpu_id_); // get max_gpu_mem and max_gpu_util std::vector gpu_tokens; - split(gpu_mem_info, gpu_tokens, ','); + Split(gpu_mem_info, gpu_tokens, ','); max_gpu_mem_ = std::max(max_gpu_mem_, stof(gpu_tokens[6])); max_gpu_util_ = std::max(max_gpu_util_, stof(gpu_tokens[7])); #endif - if (stop_signal_) break; + if (stop_signal_) { + break; + } std::this_thread::sleep_for( std::chrono::milliseconds(sampling_interval_)); } @@ -121,26 +125,6 @@ void ResourceUsageMonitor::StopInternal() { check_memory_thd_.reset(nullptr); } -std::string ResourceUsageMonitor::GetCurrentCpuMemoryInfo() { - std::string result = ""; -#if defined(__linux__) || defined(__ANDROID__) - int iPid = static_cast(getpid()); - std::string command = "pmap -x " + std::to_string(iPid) + " | grep total"; - FILE* pp = popen(command.data(), "r"); - if (!pp) return ""; - char tmp[1024]; - - while (fgets(tmp, sizeof(tmp), pp) != NULL) { - result += tmp; - } - pclose(pp); -#else - FDASSERT(false, - "Currently collect cpu memory info only supports Linux and ANDROID.") -#endif - return result; -} - std::string ResourceUsageMonitor::GetCurrentGpuMemoryInfo(int device_id) { std::string result = ""; #if defined(__linux__) && defined(WITH_GPU) diff --git a/fastdeploy/benchmark/utils.h b/fastdeploy/benchmark/utils.h index f81cb29c1..4037cd09c 100755 --- a/fastdeploy/benchmark/utils.h +++ b/fastdeploy/benchmark/utils.h @@ -65,20 +65,26 @@ class FASTDEPLOY_DECL ResourceUsageMonitor { private: void StopInternal(); - // Get current cpu memory info - std::string GetCurrentCpuMemoryInfo(); // Get current gpu memory info std::string GetCurrentGpuMemoryInfo(int device_id); bool is_supported_ = false; bool stop_signal_ = false; const int sampling_interval_; - float max_cpu_mem_ = 0.0f; - float max_gpu_mem_ = 0.0f; + float max_cpu_mem_ = 0.0f; // MB + float max_gpu_mem_ = 0.0f; // MB float max_gpu_util_ = 0.0f; const int gpu_id_ = 0; std::unique_ptr check_memory_thd_ = nullptr; }; +// Remove the ch characters at both ends of str +FASTDEPLOY_DECL std::string Strip(const std::string& str, char ch = ' '); + +// Split string +FASTDEPLOY_DECL void Split(const std::string& s, + std::vector& tokens, + char delim = ' '); + } // namespace benchmark } // namespace fastdeploy From cd5a54cd75057dfa4877005e0cee513c1b628e06 Mon Sep 17 00:00:00 2001 From: Zheng-Bicheng <58363586+Zheng-Bicheng@users.noreply.github.com> Date: Tue, 14 Feb 2023 15:40:15 +0800 Subject: [PATCH 27/30] [Bug Fix] Fix the bug reported by RK3568 Runtime under the latest driver (#1311) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * 更新rkyolo runtime 在rk3568最新驱动下报错的问题 * 删除debug代码 --- fastdeploy/runtime/backends/rknpu2/rknpu2_backend.cc | 3 ++- fastdeploy/runtime/backends/rknpu2/rknpu2_backend.h | 4 ++-- fastdeploy/runtime/runtime_option.h | 4 ++-- 3 files changed, 6 insertions(+), 5 deletions(-) diff --git a/fastdeploy/runtime/backends/rknpu2/rknpu2_backend.cc b/fastdeploy/runtime/backends/rknpu2/rknpu2_backend.cc index dc7bbbad5..a6b9a386f 100644 --- a/fastdeploy/runtime/backends/rknpu2/rknpu2_backend.cc +++ b/fastdeploy/runtime/backends/rknpu2/rknpu2_backend.cc @@ -67,6 +67,7 @@ bool RKNPU2Backend::GetSDKAndDeviceVersion() { ***************************************************************/ void RKNPU2Backend::BuildOption(const RKNPU2BackendOption& option) { this->option_ = option; + // save cpu_name this->option_.cpu_name = option.cpu_name; @@ -123,7 +124,7 @@ bool RKNPU2Backend::InitFromRKNN(const std::string& model_file, * @return bool * @note Only support RK3588 ***************************************************************/ -bool RKNPU2Backend::SetCoreMask(rknpu2::CoreMask& core_mask) const { +bool RKNPU2Backend::SetCoreMask(const rknpu2::CoreMask& core_mask) { int ret = rknn_set_core_mask(ctx, static_cast(core_mask)); if (ret != RKNN_SUCC) { FDERROR << "rknn_set_core_mask fail! ret=" << ret << std::endl; diff --git a/fastdeploy/runtime/backends/rknpu2/rknpu2_backend.h b/fastdeploy/runtime/backends/rknpu2/rknpu2_backend.h index 8b3aa9fe2..52f174d6d 100644 --- a/fastdeploy/runtime/backends/rknpu2/rknpu2_backend.h +++ b/fastdeploy/runtime/backends/rknpu2/rknpu2_backend.h @@ -25,7 +25,7 @@ namespace fastdeploy { struct RKNPU2BackendOption { - rknpu2::CpuName cpu_name = rknpu2::CpuName::RK3588; + rknpu2::CpuName cpu_name = rknpu2::CpuName::RK356X; // The specification of NPU core setting.It has the following choices : // RKNN_NPU_CORE_AUTO : Referring to automatic mode, meaning that it will @@ -49,7 +49,7 @@ class RKNPU2Backend : public BaseBackend { bool GetSDKAndDeviceVersion(); - bool SetCoreMask(rknpu2::CoreMask& core_mask) const; + bool SetCoreMask(const rknpu2::CoreMask& core_mask); bool GetModelInputOutputInfos(); diff --git a/fastdeploy/runtime/runtime_option.h b/fastdeploy/runtime/runtime_option.h index a36ac5459..904640eab 100755 --- a/fastdeploy/runtime/runtime_option.h +++ b/fastdeploy/runtime/runtime_option.h @@ -71,9 +71,9 @@ struct FASTDEPLOY_DECL RuntimeOption { void UseGpu(int gpu_id = 0); /// Use RKNPU2 e.g RK3588/RK356X to inference void UseRKNPU2(fastdeploy::rknpu2::CpuName rknpu2_name = - fastdeploy::rknpu2::CpuName::RK3588, + fastdeploy::rknpu2::CpuName::RK356X, fastdeploy::rknpu2::CoreMask rknpu2_core = - fastdeploy::rknpu2::CoreMask::RKNN_NPU_CORE_0); + fastdeploy::rknpu2::CoreMask::RKNN_NPU_CORE_AUTO); /// Use TimVX e.g RV1126/A311D to inference void UseTimVX(); /// Use Huawei Ascend to inference From fc6edcc541c98aeb28093511d419765780b0d7b9 Mon Sep 17 00:00:00 2001 From: Zheng-Bicheng <58363586+Zheng-Bicheng@users.noreply.github.com> Date: Tue, 14 Feb 2023 17:34:41 +0800 Subject: [PATCH 28/30] [Model] Update PPDetection RKNPU2 (#1323) MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit * 更新docs * 修正docs错误 * 更新docs * 更新python example脚本和ppyoloe转换脚本 --- .../paddledetection/rknpu2/README_CN.md | 81 ++++++++++++++++++- .../paddledetection/rknpu2/picodet.md | 68 ---------------- .../paddledetection/rknpu2/python/infer.py | 7 +- .../paddledetection/rknpu2/yolov8.md | 50 ------------ .../picodet_s_416_coco_lcnet_unquantized.yaml | 2 +- ...ppyoloe_plus_crn_s_80e_coco_quantized.yaml | 17 ++++ 6 files changed, 99 insertions(+), 126 deletions(-) delete mode 100644 examples/vision/detection/paddledetection/rknpu2/picodet.md delete mode 100644 examples/vision/detection/paddledetection/rknpu2/yolov8.md create mode 100644 tools/rknpu2/config/ppyoloe_plus_crn_s_80e_coco_quantized.yaml diff --git a/examples/vision/detection/paddledetection/rknpu2/README_CN.md b/examples/vision/detection/paddledetection/rknpu2/README_CN.md index 8fb3765ba..6a932fe3a 100644 --- a/examples/vision/detection/paddledetection/rknpu2/README_CN.md +++ b/examples/vision/detection/paddledetection/rknpu2/README_CN.md @@ -4,12 +4,14 @@ ## 支持模型列表 -目前FastDeploy使用RKNPU2支持如下PaddleDetection模型的部署: +在RKNPU2上已经通过测试的PaddleDetection模型如下: - Picodet -- PPYOLOE +- PPYOLOE(int8) - YOLOV8 +如果你需要查看详细的速度信息，请查看[RKNPU2模型速度一览表](../../../../../docs/cn/faq/rknpu2/rknpu2.md) + ## 准备PaddleDetection部署模型以及转换模型 RKNPU部署模型前需要将Paddle模型转换成RKNN模型，具体步骤如下: @@ -20,8 +22,79 @@ RKNPU部署模型前需要将Paddle模型转换成RKNN模型，具体步骤如 ## 模型转换example -- [Picodet RKNPU2模型转换文档](./picodet.md) -- [YOLOv8 RKNPU2模型转换文档](./yolov8.md) +### 注意点 + +PPDetection模型在RKNPU2上部署时要注意以下几点: + +* 模型导出需要包含Decode +* 由于RKNPU2不支持NMS，因此输出节点必须裁剪至NMS之前 +* 由于RKNPU2 Div算子的限制，模型的输出节点需要裁剪至Div算子之前 + +### Paddle模型转换为ONNX模型 + +由于Rockchip提供的rknn-toolkit2工具暂时不支持Paddle模型直接导出为RKNN模型，因此需要先将Paddle模型导出为ONNX模型，再将ONNX模型转为RKNN模型。 + +```bash +# 以Picodet为例 +# 下载Paddle静态图模型并解压 +wget https://paddledet.bj.bcebos.com/deploy/Inference/picodet_s_416_coco_lcnet.tar +tar xvf picodet_s_416_coco_lcnet.tar + +# 静态图转ONNX模型，注意，这里的save_file请和压缩包名对齐 +paddle2onnx --model_dir picodet_s_416_coco_lcnet \ + --model_filename model.pdmodel \ + --params_filename model.pdiparams \ + --save_file picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx \ + --enable_dev_version True + +# 固定shape +python -m paddle2onnx.optimize --input_model picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx \ + --output_model picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx \ + --input_shape_dict "{'image':[1,3,416,416]}" +``` + +### 编写yaml文件 + +**修改normalize参数** + +如果你需要在NPU上执行normalize操作，请根据你的模型配置normalize参数，例如: + +```yaml +mean: + - + - 123.675 + - 116.28 + - 103.53 +std: + - + - 58.395 + - 57.12 + - 57.375 +``` + +**修改outputs参数** +由于Paddle2ONNX版本的不同，转换模型的输出节点名称也有所不同，请使用[Netron](https://netron.app)对模型进行可视化，并找到以下蓝色方框标记的NonMaxSuppression节点，红色方框的节点名称即为目标名称。 + +例如，使用Netron可视化后，得到以下图片: + +![](https://user-images.githubusercontent.com/58363586/212599781-e1952da7-6eae-4951-8ca7-bab7e6940692.png) + +找到蓝色方框标记的NonMaxSuppression节点，可以看到红色方框标记的两个节点名称为p2o.Div.79和p2o.Concat.9,因此需要修改outputs参数，修改后如下: + +```yaml +outputs_nodes: + - 'p2o.Mul.179' + - 'p2o.Concat.9' +``` + +### ONNX模型转RKNN模型 + +为了方便大家使用，我们提供了python脚本，通过我们预配置的config文件，你将能够快速地转换ONNX模型到RKNN模型 + +```bash +python tools/rknpu2/export.py --config_path tools/rknpu2/config/picodet_s_416_coco_lcnet_unquantized.yaml \ + --target_platform rk3588 +``` ## 其他链接 diff --git a/examples/vision/detection/paddledetection/rknpu2/picodet.md b/examples/vision/detection/paddledetection/rknpu2/picodet.md deleted file mode 100644 index de0192648..000000000 --- a/examples/vision/detection/paddledetection/rknpu2/picodet.md +++ /dev/null @@ -1,68 +0,0 @@ -# Picodet RKNPU2模型转换文档 - -以下步骤均在Ubuntu电脑上完成，请参考配置文档完成转换模型环境配置。下面以Picodet-s为例子,教大家如何转换PaddleDetection模型到RKNN模型。 - - -### 导出ONNX模型 - -```bash -# 下载Paddle静态图模型并解压 -wget https://paddledet.bj.bcebos.com/deploy/Inference/picodet_s_416_coco_lcnet.tar -tar xvf picodet_s_416_coco_lcnet.tar - -# 静态图转ONNX模型，注意，这里的save_file请和压缩包名对齐 -paddle2onnx --model_dir picodet_s_416_coco_lcnet \ - --model_filename model.pdmodel \ - --params_filename model.pdiparams \ - --save_file picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx \ - --enable_dev_version True - -# 固定shape -python -m paddle2onnx.optimize --input_model picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx \ - --output_model picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx \ - --input_shape_dict "{'image':[1,3,416,416]}" -``` - -### 编写模型导出配置文件 - -以转化RK3568的RKNN模型为例子，我们需要编辑tools/rknpu2/config/picodet_s_416_coco_lcnet_unquantized.yaml，来转换ONNX模型到RKNN模型。 - -**修改normalize参数** - -如果你需要在NPU上执行normalize操作，请根据你的模型配置normalize参数，例如: - -```yaml -mean: - - - - 127.5 - - 127.5 - - 127.5 -std: - - - - 127.5 - - 127.5 - - 127.5 -``` - -**修改outputs参数** -由于Paddle2ONNX版本的不同，转换模型的输出节点名称也有所不同，请使用[Netron](https://netron.app)对模型进行可视化，并找到以下蓝色方框标记的NonMaxSuppression节点，红色方框的节点名称即为目标名称。 - -例如，使用Netron可视化后，得到以下图片: - -![](https://user-images.githubusercontent.com/58363586/212599781-e1952da7-6eae-4951-8ca7-bab7e6940692.png) - -找到蓝色方框标记的NonMaxSuppression节点，可以看到红色方框标记的两个节点名称为p2o.Div.79和p2o.Concat.9,因此需要修改outputs参数，修改后如下: - -```yaml -outputs_nodes: [ 'p2o.Div.79','p2o.Concat.9' ] -``` - -### 转换模型 - -```bash - -# ONNX模型转RKNN模型 -# 转换模型,模型将生成在picodet_s_320_coco_lcnet_non_postprocess目录下 -python tools/rknpu2/export.py --config_path tools/rknpu2/config/picodet_s_416_coco_lcnet_unquantized.yaml \ - --target_platform rk3588 -``` diff --git a/examples/vision/detection/paddledetection/rknpu2/python/infer.py b/examples/vision/detection/paddledetection/rknpu2/python/infer.py index a3c146531..2dfb54281 100644 --- a/examples/vision/detection/paddledetection/rknpu2/python/infer.py +++ b/examples/vision/detection/paddledetection/rknpu2/python/infer.py @@ -45,15 +45,16 @@ if __name__ == "__main__": # 配置runtime，加载模型 runtime_option = fd.RuntimeOption() - runtime_option.use_cpu() + runtime_option.use_rknpu2() model = fd.vision.detection.PPYOLOE( model_file, params_file, config_file, runtime_option=runtime_option, - model_format=fd.ModelFormat.ONNX) - + model_format=fd.ModelFormat.RKNN) + model.preprocessor.disable_normalize() + model.preprocessor.disable_permute() model.postprocessor.apply_decode_and_nms() # 预测图片分割结果 diff --git a/examples/vision/detection/paddledetection/rknpu2/yolov8.md b/examples/vision/detection/paddledetection/rknpu2/yolov8.md deleted file mode 100644 index 432fe02bb..000000000 --- a/examples/vision/detection/paddledetection/rknpu2/yolov8.md +++ /dev/null @@ -1,50 +0,0 @@ -# YOLOv8 RKNPU2模型转换文档 - -以下步骤均在Ubuntu电脑上完成，请参考配置文档完成转换模型环境配置。下面以yolov8为例子,教大家如何转换PaddleDetection模型到RKNN模型。 - - -### 导出ONNX模型 - -```bash -# 下载Paddle静态图模型并解压 - -# 静态图转ONNX模型，注意，这里的save_file请和压缩包名对齐 -paddle2onnx --model_dir yolov8_n_500e_coco \ - --model_filename model.pdmodel \ - --params_filename model.pdiparams \ - --save_file yolov8_n_500e_coco/yolov8_n_500e_coco.onnx \ - --enable_dev_version True - -# 固定shape -python -m paddle2onnx.optimize --input_model yolov8_n_500e_coco/yolov8_n_500e_coco.onnx \ - --output_model yolov8_n_500e_coco/yolov8_n_500e_coco.onnx \ - --input_shape_dict "{'image':[1,3,640,640],'scale_factor':[1,2]}" -``` - -### 编写模型导出配置文件 -**修改outputs参数** -由于Paddle2ONNX版本的不同，转换模型的输出节点名称也有所不同，请使用[Netron](https://netron.app)对模型进行可视化，并找到以下蓝色方框标记的NonMaxSuppression节点，红色方框的节点名称即为目标名称。 - -例如，使用Netron可视化后，得到以下图片: - -![](https://user-images.githubusercontent.com/58363586/212599658-8a2c4b79-f59a-40b5-ade7-f77c6fcfdf2a.png) - -找到蓝色方框标记的NonMaxSuppression节点，可以看到红色方框标记的两个节点名称为p2o.Div.1和p2o.Concat.9,因此需要修改outputs参数，修改后如下: - -```yaml -outputs_nodes: [ 'p2o.Div.1','p2o.Concat.49' ] -``` - -### 转换模型 - -```bash - -# ONNX模型转RKNN模型 -# 转换非全量化模型,模型将生成在yolov8_n目录下 -python tools/rknpu2/export.py --config_path tools/rknpu2/config/yolov8_n_unquantized.yaml \ - --target_platform rk3588 - -# 转换全量化模型,模型将生成在yolov8_n目录下 -python tools/rknpu2/export.py --config_path tools/rknpu2/config/yolov8_n_quantized.yaml \ - --target_platform rk3588 -``` diff --git a/tools/rknpu2/config/picodet_s_416_coco_lcnet_unquantized.yaml b/tools/rknpu2/config/picodet_s_416_coco_lcnet_unquantized.yaml index aeeb3fdb7..ddcd9c91e 100644 --- a/tools/rknpu2/config/picodet_s_416_coco_lcnet_unquantized.yaml +++ b/tools/rknpu2/config/picodet_s_416_coco_lcnet_unquantized.yaml @@ -10,7 +10,7 @@ std: - 57.375 model_path: ./picodet_s_416_coco_lcnet/picodet_s_416_coco_lcnet.onnx outputs_nodes: - - 'p2o.Div.79' + - 'p2o.Mul.179' - 'p2o.Concat.9' do_quantization: False dataset: diff --git a/tools/rknpu2/config/ppyoloe_plus_crn_s_80e_coco_quantized.yaml b/tools/rknpu2/config/ppyoloe_plus_crn_s_80e_coco_quantized.yaml new file mode 100644 index 000000000..965e43cde --- /dev/null +++ b/tools/rknpu2/config/ppyoloe_plus_crn_s_80e_coco_quantized.yaml @@ -0,0 +1,17 @@ +mean: + - + - 0 + - 0 + - 0 +std: + - + - 255 + - 255 + - 255 +model_path: ./ppyoloe_plus_crn_s_80e_coco/ppyoloe_plus_crn_s_80e_coco.onnx +outputs_nodes: + - 'p2o.Mul.224' + - 'p2o.Concat.29' +do_quantization: True +dataset: "./ppyoloe_plus_crn_s_80e_coco/dataset.txt" +output_folder: "./ppyoloe_plus_crn_s_80e_coco" From c25d1cc1bce44ef56fc7c86ae9bd7f3f931b9b65 Mon Sep 17 00:00:00 2001 From: WJJ1995 Date: Tue, 14 Feb 2023 17:51:39 +0800 Subject: [PATCH 29/30] [Backend]Fixed enable_paddle_to_trt() bug (#1320) * add GPL lisence * add GPL-3.0 lisence * add GPL-3.0 lisence * add GPL-3.0 lisence * support yolov8 * add pybind for yolov8 * add yolov8 readme * add cpp benchmark * add cpu and gpu mem * public part split * add runtime mode * fixed bugs * add cpu_thread_nums * deal with comments * deal with comments * deal with comments * rm useless code * add FASTDEPLOY_DECL * add FASTDEPLOY_DECL * fixed for windows * mv rss to pss * mv rss to pss * Update utils.cc * use thread to collect mem * Add ResourceUsageMonitor * rm useless code * fixed bug * fixed typo * update ResourceUsageMonitor * fixed bug * fixed bug * add note for ResourceUsageMonitor * deal with comments * add macros * deal with comments * deal with comments * deal with comments * re-lint * rm pmap and use mem api * rm pmap and use mem api * add mem api * Add PrintBenchmarkInfo func * Add PrintBenchmarkInfo func * Add PrintBenchmarkInfo func * deal with comments * fixed enable_paddle_to_trt * add log for paddle_trt --------- Co-authored-by: DefTruth <31974251+DefTruth@users.noreply.github.com> --- fastdeploy/runtime/backends/paddle/paddle_backend.cc | 4 ++++ python/fastdeploy/runtime.py | 5 +++-- 2 files changed, 7 insertions(+), 2 deletions(-) mode change 100644 => 100755 fastdeploy/runtime/backends/paddle/paddle_backend.cc mode change 100644 => 100755 python/fastdeploy/runtime.py diff --git a/fastdeploy/runtime/backends/paddle/paddle_backend.cc b/fastdeploy/runtime/backends/paddle/paddle_backend.cc old mode 100644 new mode 100755 index 09dbe812a..dc804e926 --- a/fastdeploy/runtime/backends/paddle/paddle_backend.cc +++ b/fastdeploy/runtime/backends/paddle/paddle_backend.cc @@ -29,6 +29,10 @@ void PaddleBackend::BuildOption(const PaddleBackendOption& option) { config_.SetExecStream(option_.external_stream_); } if (option.enable_trt) { + if (!option.trt_option.enable_fp16) { + FDINFO << "Will try to use tensorrt inference with Paddle Backend." + << std::endl; + } config_.Exp_DisableTensorRtOPs(option.trt_disabled_ops_); auto precision = paddle_infer::PrecisionType::kFloat32; if (option.trt_option.enable_fp16) { diff --git a/python/fastdeploy/runtime.py b/python/fastdeploy/runtime.py old mode 100644 new mode 100755 index 6be764ea3..4980db52e --- a/python/fastdeploy/runtime.py +++ b/python/fastdeploy/runtime.py @@ -532,9 +532,10 @@ class RuntimeOption: logging.warning(" option = fd.RuntimeOption()") logging.warning(" option.use_gpu(0)") logging.warning(" option.use_paddle_infer_backend()") - logging.warning(" option.paddle_infer_option.enabel_trt = True") + logging.warning(" option.paddle_infer_option.enable_trt = True") logging.warning(" ==============================================") - return self._option.enable_paddle_to_trt() + self._option.use_paddle_backend() + self._option.paddle_infer_option.enable_trt = True def set_trt_max_workspace_size(self, trt_max_workspace_size): """Set max workspace size while using TensorRT backend. From a5d23c57d0cb6bd52fd2bc3daa4a9d3c2274c403 Mon Sep 17 00:00:00 2001 From: CoolCola <49013063+CoolKbh@users.noreply.github.com> Date: Tue, 14 Feb 2023 18:36:28 +0800 Subject: [PATCH 30/30] [Bug fix]add yolov7face landmarks (#1297) * add yolov7face benchmark * fix review problem * fix review problems --- .../contrib/yolov7face/postprocessor.cc | 27 ++++++++++++++++--- .../contrib/yolov7face/postprocessor.h | 12 +++++++-- .../contrib/yolov7face/yolov7face_pybind.cc | 3 ++- .../vision/facedet/contrib/yolov7face.py | 12 +++++++++ 4 files changed, 48 insertions(+), 6 deletions(-) diff --git a/fastdeploy/vision/facedet/contrib/yolov7face/postprocessor.cc b/fastdeploy/vision/facedet/contrib/yolov7face/postprocessor.cc index 30bb523cc..624aa3403 100644 --- a/fastdeploy/vision/facedet/contrib/yolov7face/postprocessor.cc +++ b/fastdeploy/vision/facedet/contrib/yolov7face/postprocessor.cc @@ -24,7 +24,7 @@ namespace facedet { Yolov7FacePostprocessor::Yolov7FacePostprocessor() { conf_threshold_ = 0.5; nms_threshold_ = 0.45; - max_wh_ = 7680.0; + landmarks_per_face_ = 5; } bool Yolov7FacePostprocessor::Run(const std::vector& infer_result, @@ -36,6 +36,8 @@ bool Yolov7FacePostprocessor::Run(const std::vector& infer_result, for (size_t bs = 0; bs < batch; ++bs) { (*results)[bs].Clear(); + // must be setup landmarks_per_face before reserve + (*results)[bs].landmarks_per_face = landmarks_per_face_; (*results)[bs].Reserve(infer_result[0].shape[1]); if (infer_result[0].dtype != FDDataType::FP32) { FDERROR << "Only support post process with float32 data." << std::endl; @@ -61,6 +63,15 @@ bool Yolov7FacePostprocessor::Run(const std::vector& infer_result, (*results)[bs].boxes.emplace_back(std::array{ (x - w / 2.f), (y - h / 2.f), (x + w / 2.f), (y + h / 2.f)}); (*results)[bs].scores.push_back(confidence); + + // decode landmarks (default 5 landmarks) + if (landmarks_per_face_ > 0) { + float* landmarks_ptr = const_cast(reg_cls_ptr + 6); + for (size_t j = 0; j < landmarks_per_face_ * 3; j += 3) { + (*results)[bs].landmarks.emplace_back( + std::array{landmarks_ptr[j], landmarks_ptr[j + 1]}); + } + } } if ((*results)[bs].boxes.size() == 0) { @@ -79,9 +90,9 @@ bool Yolov7FacePostprocessor::Run(const std::vector& infer_result, float ipt_h = iter_ipt->second[0]; float ipt_w = iter_ipt->second[1]; float scale = std::min(out_h / ipt_h, out_w / ipt_w); + float pad_h = (out_h - ipt_h * scale) / 2; + float pad_w = (out_w - ipt_w * scale) / 2; for (size_t i = 0; i < (*results)[bs].boxes.size(); ++i) { - float pad_h = (out_h - ipt_h * scale) / 2; - float pad_w = (out_w - ipt_w * scale) / 2; // clip box (*results)[bs].boxes[i][0] = std::max(((*results)[bs].boxes[i][0] - pad_w) / scale, 0.0f); (*results)[bs].boxes[i][1] = std::max(((*results)[bs].boxes[i][1] - pad_h) / scale, 0.0f); @@ -92,6 +103,16 @@ bool Yolov7FacePostprocessor::Run(const std::vector& infer_result, (*results)[bs].boxes[i][2] = std::min((*results)[bs].boxes[i][2], ipt_w - 1.0f); (*results)[bs].boxes[i][3] = std::min((*results)[bs].boxes[i][3], ipt_h - 1.0f); } + + // scale and clip landmarks + for (size_t i = 0; i < (*results)[bs].landmarks.size(); ++i) { + (*results)[bs].landmarks[i][0] = + std::max(((*results)[bs].landmarks[i][0] - pad_w) / scale, 0.0f); + (*results)[bs].landmarks[i][1] = + std::max(((*results)[bs].landmarks[i][1] - pad_h) / scale, 0.0f); + (*results)[bs].landmarks[i][0] = std::min((*results)[bs].landmarks[i][0], ipt_w - 1.0f); + (*results)[bs].landmarks[i][1] = std::min((*results)[bs].landmarks[i][1], ipt_h - 1.0f); + } } return true; } diff --git a/fastdeploy/vision/facedet/contrib/yolov7face/postprocessor.h b/fastdeploy/vision/facedet/contrib/yolov7face/postprocessor.h index 4a1b0d852..f6a6bb23c 100644 --- a/fastdeploy/vision/facedet/contrib/yolov7face/postprocessor.h +++ b/fastdeploy/vision/facedet/contrib/yolov7face/postprocessor.h @@ -56,11 +56,19 @@ class FASTDEPLOY_DECL Yolov7FacePostprocessor{ /// Get nms_threshold, default 0.45 float GetNMSThreshold() const { return nms_threshold_; } + /// Set landmarks_per_face, default 5 + void SetLandmarksPerFace(const int& landmarks_per_face) { + landmarks_per_face_ = landmarks_per_face; + } + + /// Get landmarks_per_face, default 5 + int GetLandmarksPerFace() const { return landmarks_per_face_; } + + protected: float conf_threshold_; float nms_threshold_; - bool multi_label_; - float max_wh_; + int landmarks_per_face_; }; } // namespace facedet diff --git a/fastdeploy/vision/facedet/contrib/yolov7face/yolov7face_pybind.cc b/fastdeploy/vision/facedet/contrib/yolov7face/yolov7face_pybind.cc index c0c99d425..07f5feece 100644 --- a/fastdeploy/vision/facedet/contrib/yolov7face/yolov7face_pybind.cc +++ b/fastdeploy/vision/facedet/contrib/yolov7face/yolov7face_pybind.cc @@ -60,7 +60,8 @@ void BindYOLOv7Face(pybind11::module& m) { return results; }) .def_property("conf_threshold", &vision::facedet::Yolov7FacePostprocessor::GetConfThreshold, &vision::facedet::Yolov7FacePostprocessor::SetConfThreshold) - .def_property("nms_threshold", &vision::facedet::Yolov7FacePostprocessor::GetNMSThreshold, &vision::facedet::Yolov7FacePostprocessor::SetNMSThreshold); + .def_property("nms_threshold", &vision::facedet::Yolov7FacePostprocessor::GetNMSThreshold, &vision::facedet::Yolov7FacePostprocessor::SetNMSThreshold) + .def_property("landmarks_per_face", &vision::facedet::Yolov7FacePostprocessor::GetLandmarksPerFace, &vision::facedet::Yolov7FacePostprocessor::SetLandmarksPerFace); pybind11::class_(m, "YOLOv7Face") .def(pybind11::init