add face detect wrid face

cfg/face.data

+cfg_model=yolo
+classes=1
+gpus = 0
+num_workers = 8
+batch_size = 8
+img_size = 416
+multi_scale = True
+epochs = 100
+train=./yolo_widerface_open_train/anno/train.txt
+valid=./yolo_widerface_open_train/anno/train.txt
+names=./cfg/face.names
+#finetune_model=./coco_model/yolov3_coco.pt
+finetune_model = ./weights-yolov3-face/latest_416.pt
+lr_step = 10,20,30
+lr0 = 0.0001

cfg/face.names

+Face

predict.py

@@ -55,7 +55,6 @@ def refine_hand_bbox(bbox,img_shape):
     return (x1,y1,x2,y2)
 def detect(
         model_path,
-        root_path,
         cfg,
         data_cfg,
         img_size=416,
@@ -94,12 +93,20 @@ def detect(
 
     colors = [(v // 32 * 64 + 64, (v // 8) % 4 * 64, v % 8 * 32) for v in range(1, num_classes + 1)][::-1]
 
-    video_capture = cv2.VideoCapture(0)
+    video_capture = cv2.VideoCapture("./video/bean.mp4")
+    # url="http://admin:admin@192.168.43.1:8081"
+    # video_capture=cv2.VideoCapture(url)
+
+    video_writer = None
+    loc_time = time.localtime()
+    str_time = time.strftime("%Y-%m-%d_%H-%M-%S", loc_time)
+    save_video_path = "./demo/demo_{}.mp4".format(str_time)
     #-------------------------------------------------
     while True:
         ret, im0 = video_capture.read()
         if ret:
             t = time.time()
+            # im0 = cv2.imread("samples/5.png")
             img = process_data(im0, img_size)
             if use_cuda:
                 torch.cuda.synchronize()
@@ -140,8 +147,11 @@ def detect(
             # Draw bounding boxes and labels of detections
             for *xyxy, conf, cls_conf, cls in detections:
                 label = '%s %.2f' % (classes[int(cls)], conf)
+                # label = '%s' % (classes[int(cls)])
+
+                # print(conf, cls_conf)
                 # xyxy = refine_hand_bbox(xyxy,im0.shape)
-                plot_one_box(xyxy, im0, label=label, color=(255,255,0))
+                plot_one_box(xyxy, im0, label=label, color=(155,55,255),line_thickness = 3)
 
             s2 = time.time()
             print("detect time: {} \n".format(s2 - t))
@@ -153,33 +163,31 @@ def detect(
             cv2.namedWindow('image',0)
             cv2.imshow("image", im0)
             key = cv2.waitKey(1)
+            if video_writer is None:
+                fourcc = cv2.VideoWriter_fourcc(*"mp4v")
+                video_writer = cv2.VideoWriter(save_video_path, fourcc, fps=25, frameSize = (im0.shape[1], im0.shape[0]))
+            video_writer.write(im0)
             if key == 27:
                 break
         else:
             break
 
     cv2.destroyAllWindows()
+    video_writer.release()
 
 if __name__ == '__main__':
-    pattern = 'yolo'
-    if "-tiny" in pattern:
-        model_path = './weights-yolov3-person-tiny/latest_320.pt' # 检测模型路径
-        root_path = './test_images/'# 测试文件夹
-        model_cfg = pattern # 模型类型
-    else:
-        model_path = './weights-yolov3-hand/latest_416.pt' # 检测模型路径
-        root_path = './test_images/'# 测试文件夹
-        model_cfg = 'yolov3' # 模型类型
 
-    voc_config = 'cfg/hand.data' # 模型相关配置文件
+    voc_config = 'cfg/face.data' # 模型相关配置文件
+    model_path = './weights-yolov3-face/latest_416.pt' # 检测模型路径
+    model_cfg = 'yolo' # yolo / yolo-tiny
+
     img_size = 416 # 图像尺寸
-    conf_thres = 0.25# 检测置信度
-    nms_thres = 0.45 # nms 阈值
+    conf_thres = 0.36# 检测置信度
+    nms_thres = 0.5 # nms 阈值
 
     with torch.no_grad():#设置无梯度运行
         detect(
             model_path = model_path,
-            root_path = root_path,
             cfg = model_cfg,
             data_cfg = voc_config,
             img_size=img_size,

train.py

@@ -202,7 +202,8 @@ def train(data_cfg ='cfg/voc.data',
 if __name__ == '__main__':
 
 
-    train(data_cfg='cfg/hand.data')
+    # train(data_cfg="cfg/hand.data")
+    train(data_cfg = "cfg/face.data")
 
 
     print('well done ~ ')