need to deal with image

.vscode/settings.json

@@ -0,0 +1,6 @@
+{
+    "files.associations": {
+        "detection.h": "c",
+        "network.h": "c"
+    }
+}

classes.c

@@ -0,0 +1,8 @@
+#include <stdlib.h>
+
+char * get_class_name(char **names, int index, int names_len) {
+    if (index >= names_len) {
+        return NULL;
+    }
+    return names[index];
+}

classes.go

@@ -0,0 +1,15 @@
+package darknet
+
+// #include <stdlib.h>
+// #include "classes.h"
+import "C"
+
+func makeClassNames(names **C.char, classes int) []string {
+	out := make([]string, classes)
+	for i := 0; i < classes; i++ {
+		n := C.get_class_name(names, C.int(i), C.int(classes))
+		s := C.GoString(n)
+		out[i] = s
+	}
+	return out
+}

classes.h

@@ -0,0 +1,3 @@
+#pragma once
+
+extern char * get_class_name(char **names, int index, int names_len);

darknet.go

@@ -1,5 +1,5 @@
 package darknet
 
 // #cgo CFLAGS: -I/usr/local/include -I/usr/local/cuda/include
-// #cgo LDFLAGS: -L./lib -ldark -lm
+// #cgo LDFLAGS: -L./lib -ldarknet -lm
 import "C"

data/names.list

@@ -0,0 +1,5 @@
+person
+car
+motorcycle
+bus
+truck

detection.c

@@ -0,0 +1,22 @@
+#include <darknet.h>
+
+#include "detection.h"
+
+
+detection * get_detection(detection *dets, int index, int dets_len)
+{
+    if (index >= dets_len) {
+        return NULL;
+    }
+
+    return dets + index;
+}
+
+float get_detection_probability(detection *det, int index, int prob_len)
+{
+    if (index >= prob_len) {
+        return .0;
+    }
+
+    return det->prob[index];
+}

detection.go

@@ -1,10 +1,81 @@
 package darknet
 
-import "time"
+// #include <darknet.h>
+//
+// #include "detection.h"
+import "C"
+import (
+	"image"
+	"time"
+)
+
+// Detection represents a detection.
+type Detection struct {
+	BoundingBox
+
+	ClassIDs      []int
+	ClassNames    []string
+	Probabilities []float32
+}
 
 // DetectionResult represents the inference results from the network.
 type DetectionResult struct {
-	// Detections           []*Detection
+	Detections           []*Detection
 	NetworkOnlyTimeTaken time.Duration
 	OverallTimeTaken     time.Duration
 }
+
+func makeDetection(img *DarknetImage, det *C.detection, threshold float32, classes int, classNames []string) *Detection {
+
+	if det == nil {
+		return &Detection{}
+	}
+
+	dClassIDs := make([]int, 0)
+	dClassNames := make([]string, 0)
+	dProbs := make([]float32, 0)
+
+	for i := 0; i < int(classes); i++ {
+		dProb := float32(C.get_detection_probability(det, C.int(i), C.int(classes)))
+		if dProb > threshold {
+			dClassIDs = append(dClassIDs, i)
+			cN := classNames[i]
+			dClassNames = append(dClassNames, cN)
+			dProbs = append(dProbs, dProb*100)
+		}
+	}
+
+	fImgW := C.float(img.Width)
+	fImgH := C.float(img.Height)
+	halfRatioW := det.bbox.w / 2.0
+	halfRatioH := det.bbox.h / 2.0
+	out := Detection{
+		BoundingBox: BoundingBox{
+			StartPoint: image.Point{
+				X: int((det.bbox.x - halfRatioW) * fImgW),
+				Y: int((det.bbox.y - halfRatioH) * fImgH),
+			},
+			EndPoint: image.Point{
+				X: int((det.bbox.x + halfRatioW) * fImgW),
+				Y: int((det.bbox.y + halfRatioH) * fImgH),
+			},
+		},
+
+		ClassIDs:      dClassIDs,
+		ClassNames:    dClassNames,
+		Probabilities: dProbs,
+	}
+
+	return &out
+}
+
+func makeDetections(img *DarknetImage, detections *C.detection, detectionsLength int, threshold float32, classes int, classNames []string) []*Detection {
+	// Make list of detection objects.
+	ds := make([]*Detection, detectionsLength)
+	for i := 0; i < int(detectionsLength); i++ {
+		det := C.get_detection(detections, C.int(i), C.int(detectionsLength))
+		d := makeDetection(img, det, threshold, classes, classNames)
+		ds[i] = d
+	}
+	return ds
+}

detection.h

@@ -0,0 +1,7 @@
+#pragma once
+
+#include <darknet.h>
+
+
+extern detection * get_detection(detection *dets, int index, int dets_len);
+extern float get_detection_probability(detection *det, int index, int prob_len);

example/main.go

@@ -2,6 +2,7 @@ package main
 
 import (
 	"flag"
+	"fmt"
 	"image/jpeg"
 	"log"
 	"os"
@@ -37,7 +38,7 @@ func main() {
 		DataConfigurationFile:    *dataConfigFile,
 		NetworkConfigurationFile: *configFile,
 		WeightsFile:              *weightsFile,
-		Threshold:                .5,
+		Threshold:                .25,
 	}
 	if err := n.Init(); err != nil {
 		printError(err)
@@ -54,6 +55,7 @@ func main() {
 	if err != nil {
 		panic(err.Error())
 	}
+	_ = src
 
 	imgDarknet, err := darknet.Image2Float32(src)
 	if err != nil {
@@ -67,18 +69,18 @@ func main() {
 		return
 	}
 	_ = dr
-	// log.Println("Network-only time taken:", dr.NetworkOnlyTimeTaken)
-	// log.Println("Overall time taken:", dr.OverallTimeTaken, len(dr.Detections))
-	// for _, d := range dr.Detections {
 
-	// 	for i := range d.ClassIDs {
-	// 		bBox := d.BoundingBox
-	// 		fmt.Printf("%s (%d): %.4f%% | start point: (%d,%d) | end point: (%d, %d)\n",
-	// 			d.ClassNames[i], d.ClassIDs[i],
-	// 			d.Probabilities[i],
-	// 			bBox.StartPoint.X, bBox.StartPoint.Y,
-	// 			bBox.EndPoint.X, bBox.EndPoint.Y,
-	// 		)
-	// 	}
-	// }
+	log.Println("Network-only time taken:", dr.NetworkOnlyTimeTaken)
+	log.Println("Overall time taken:", dr.OverallTimeTaken, len(dr.Detections))
+	for _, d := range dr.Detections {
+		for i := range d.ClassIDs {
+			bBox := d.BoundingBox
+			fmt.Printf("%s (%d): %.4f%% | start point: (%d,%d) | end point: (%d, %d)\n",
+				d.ClassNames[i], d.ClassIDs[i],
+				d.Probabilities[i],
+				bBox.StartPoint.X, bBox.StartPoint.Y,
+				bBox.EndPoint.X, bBox.EndPoint.Y,
+			)
+		}
+	}
 }

example/out-sample.png

@@ -0,0 +1,18 @@
+[
+{
+ "frame_id":1, 
+ "filename":"/home/dimitrii/Downloads/mega.jpg", 
+ "objects": [ 
+  {"class_id":3, "name":"bus", "relative_coordinates":{"center_x":0.475497, "center_y":0.317034, "width":0.175112, "height":0.104746}, "confidence":0.686032}, 
+  {"class_id":3, "name":"bus", "relative_coordinates":{"center_x":0.278355, "center_y":0.302800, "width":0.087061, "height":0.052067}, "confidence":0.562420}, 
+  {"class_id":1, "name":"car", "relative_coordinates":{"center_x":0.799558, "center_y":0.643484, "width":0.388328, "height":0.539759}, "confidence":0.942335}, 
+  {"class_id":1, "name":"car", "relative_coordinates":{"center_x":0.365836, "center_y":0.431383, "width":0.267126, "height":0.156932}, "confidence":0.776083}, 
+  {"class_id":1, "name":"car", "relative_coordinates":{"center_x":0.513763, "center_y":0.475799, "width":0.301966, "height":0.240549}, "confidence":0.684380}, 
+  {"class_id":1, "name":"car", "relative_coordinates":{"center_x":0.717148, "center_y":0.516788, "width":0.456320, "height":0.313704}, "confidence":0.584288}, 
+  {"class_id":1, "name":"car", "relative_coordinates":{"center_x":0.217955, "center_y":0.379872, "width":0.212185, "height":0.101100}, "confidence":0.468769}, 
+  {"class_id":1, "name":"car", "relative_coordinates":{"center_x":0.126970, "center_y":0.337752, "width":0.035609, "height":0.022990}, "confidence":0.452899}, 
+  {"class_id":0, "name":"person", "relative_coordinates":{"center_x":0.051506, "center_y":0.369395, "width":0.056224, "height":0.175025}, "confidence":0.978385}, 
+  {"class_id":0, "name":"person", "relative_coordinates":{"center_x":0.097309, "center_y":0.370574, "width":0.048715, "height":0.142128}, "confidence":0.801452}
+ ] 
+}
+]

image.c

@@ -14,4 +14,9 @@ image prepare_image(image img, int w, int h, int c){
     img.h=h;
     img.c=c;
     return img;
-}
+}
+
+image new_image(int w, int h, int c){
+    image out = make_image(w, h, c);
+    return out;
+}

image.go

@@ -46,12 +46,14 @@ func Image2Float32(img image.Image) (DarknetImage, error) {
 	imgDarknet := DarknetImage{
 		Width:  width,
 		Height: height,
-		image:  C.new_darknet_image(),
+		// image:  C.new_darknet_image(),
 	}
 
-	imgDarknet.image = C.prepare_image(imgDarknet.image, C.int(width), C.int(height), 3)
-	imgDarknet.image.data = float_p(ans)
+	// imgDarknet.image = C.prepare_image(imgDarknet.image, C.int(width), C.int(height), 3)
+	// imgDarknet.image.data = float_p(ans)
 	// imgDarknet.image = C.resize_image(imgDarknet.image, 416, 416) // Do we need resize? (detection function actually does it)
 
+	imgDarknet.image = C.load_image_color(C.CString("/home/dimitrii/Downloads/mega.jpg"), 416, 416)
+
 	return imgDarknet, nil
 }

image.h

@@ -3,4 +3,5 @@
 #include <darknet.h>
 
 extern image new_darknet_image();
-extern image prepare_image(image img, int w, int h, int c);
+extern image prepare_image(image img, int w, int h, int c);
+extern image new_image(int w, int h, int c);

network.c

@@ -4,19 +4,19 @@
 
 #include "network.h"
 
-// int get_network_layer_classes(network *n, int index) {
-// 	return n->layers[index].classes;
-// }
+#include "detection.h"
 
 struct network_box_result perform_network_detect(network *n, image *img, int classes, float thresh, float hier_thresh, float nms) {
     image sized = letterbox_image(*img, n->w, n->h);
     struct network_box_result result = { NULL };
     float *X = sized.data;
-    network_predict(*n, X);
-    result.detections = get_network_boxes(n, img->w, img->h,thresh, hier_thresh, 0, 1, &result.detections_len, 0);
+    int letter_box = 0;
+    network_predict_ptr(n, X);
+    result.detections = get_network_boxes(n, img->w, img->h, thresh, hier_thresh, 0, 1, &result.detections_len, letter_box);
     if (nms) {
         do_nms_sort(result.detections, result.detections_len, classes, nms);
     }
     free_image(sized);
     return result;
 }
+

network.go

@@ -41,7 +41,18 @@ func (n *YOLONetwork) Init() error {
 	wFile := C.CString(n.WeightsFile)
 	defer C.free(unsafe.Pointer(wFile))
 
-	// GPU device ID must be set before `load_network()` is invoked.
+	// data := C.CString(n.DataConfigurationFile)
+	// defer C.free(unsafe.Pointer(data))
+
+	// inptim := C.CString("/home/dimitrii/Downloads/mega.jpg")
+	// defer C.free(unsafe.Pointer(inptim))
+
+	// outputim := C.CString("/home/dimitrii/work/src/github.com/LdDl/go-darknet/example/out-sample.png")
+	// defer C.free(unsafe.Pointer(outputim))
+
+	// C.test_detector(data, nCfg, wFile, inptim, 0.4, 0.5, 1, 1, 0, outputim, 0, 0)
+
+	// // GPU device ID must be set before `load_network()` is invoked.
 	C.cuda_set_device(C.int(n.GPUDeviceIndex))
 	n.cNet = C.load_network(nCfg, wFile, 0)
 
@@ -56,12 +67,11 @@ func (n *YOLONetwork) Init() error {
 	n.hierarchalThreshold = .5
 
 	// Currently NMS is always 0.4.
-	n.nms = .4
+	n.nms = .45
 
-	// n.Classes = int(C.get_network_layer_classes(n.cNet, n.cNet.n-1))
-	// cClassNames := loadClassNames(n.DataConfigurationFile)
-	// defer freeClassNames(cClassNames)
-	// n.ClassNames = makeClassNames(cClassNames, n.Classes)
+	metadata := C.get_metadata(nCfg)
+	n.Classes = int(metadata.classes)
+	n.ClassNames = makeClassNames(metadata.names, n.Classes)
 
 	return nil
 }
@@ -77,28 +87,21 @@ func (n *YOLONetwork) Close() error {
 	return nil
 }
 
-// Detect specified image.
+// Detect specified image
 func (n *YOLONetwork) Detect(img *DarknetImage) (*DetectionResult, error) {
 	if n.cNet == nil {
 		return nil, errNetworkNotInit
 	}
-
 	startTime := time.Now()
-	C.network_predict_image(n.cNet, img.image)
-	// result := C.perform_network_detect(n.cNet, &img.image, C.int(n.Classes), C.float(n.Threshold), C.float(n.hierarchalThreshold), C.float(n.nms))
+	result := C.perform_network_detect(n.cNet, &img.image, C.int(n.Classes), C.float(n.Threshold), C.float(n.hierarchalThreshold), C.float(n.nms))
 	endTime := time.Now()
-	// defer C.free_detections(result.detections, result.detections_len)
-
-	// ds := makeDetections(img, result.detections, int(result.detections_len),
-	// 	n.Threshold, n.Classes, n.ClassNames)
-
+	defer C.free_detections(result.detections, result.detections_len)
+	ds := makeDetections(img, result.detections, int(result.detections_len), n.Threshold, n.Classes, n.ClassNames)
 	endTimeOverall := time.Now()
-
 	out := DetectionResult{
-		// Detections:           ds,
+		Detections:           ds,
 		NetworkOnlyTimeTaken: endTime.Sub(startTime),
 		OverallTimeTaken:     endTimeOverall.Sub(startTime),
 	}
-
 	return &out, nil
 }

network.h

@@ -8,4 +8,4 @@ struct network_box_result {
 };
 
 extern int get_network_layer_classes(network *n, int index);
-extern struct network_box_result perform_network_detect(network *n, image *img, int classes, float thresh, float hier_thresh, float nms);
+extern struct network_box_result perform_network_detect(network *n, image *img, int classes, float thresh, float hier_thresh, float nms);