support show detection result, support get video resolution automatically

demo/mask_detection/python/README.md

@@ -53,8 +53,8 @@ python infer.py --models_dir=/path/to/models --img_paths=/path/to/images --video
 | models_dir | Yes|上述导出的模型路径 |
 | img_paths |img_paths/video_path 二选一|需要预测的图片目录 |
 | video_path |img_paths/video_path 二选一|需要预测的视频目录|
-| video_size |No|预测视频分辨率大小(w,h) |
-| use_camera |No|是否打开摄像头进行预测 |
+| use_camera |No|是否打开摄像头进行预测，默认为False |
+| open_imshow |No|是否进行检测结果实时绘图，默认为False |
 | use_gpu |No|是否GPU，默认为False|
 
 说明：

demo/mask_detection/python/infer.py

@@ -33,30 +33,27 @@ import argparse
 
 def parse_args():
     parser = argparse.ArgumentParser('mask detection.')
-    parser.add_argument('--models_dir',
-                        type=str,
-                        default='',
-                        help='path of models.')
-    parser.add_argument('--img_paths',
-                        type=str,
-                        default='',
-                        help='path of images')
-    parser.add_argument('--video_path',
-                        type=str,
-                        default='',
-                        help='path of video.')
-    parser.add_argument('--video_size',
-                        type=tuple,
-                        default=(1920, 1080),
-                        help='size of video.')
-    parser.add_argument('--use_camera',
-                        type=bool,
-                        default=False,
-                        help='switch detect video or camera, default:video.')
-    parser.add_argument('--use_gpu',
-                        type=bool,
-                        default=False,
-                        help='switch cpu/gpu, default:cpu.')
+    parser.add_argument(
+        '--models_dir', type=str, default='', help='path of models.')
+    parser.add_argument(
+        '--img_paths', type=str, default='', help='path of images')
+    parser.add_argument(
+        '--video_path', type=str, default='', help='path of video.')
+    parser.add_argument(
+        '--use_camera',
+        type=bool,
+        default=False,
+        help='switch detect video or camera, default:video.')
+    parser.add_argument(
+        '--open_imshow',
+        type=bool,
+        default=False,
+        help='visualize results in real time.')
+    parser.add_argument(
+        '--use_gpu',
+        type=bool,
+        default=False,
+        help='switch cpu/gpu, default:cpu.')
     args = parser.parse_args()
     return args
 
@@ -108,10 +105,11 @@ class MaskClassifier:
         h, w = self.EVAL_SIZE[1], self.EVAL_SIZE[0]
         inputs = []
         for face in faces:
-            im = cv2.resize(face.rect_data, (128, 128),
-                            fx=0,
-                            fy=0,
-                            interpolation=cv2.INTER_CUBIC)
+            im = cv2.resize(
+                face.rect_data, (128, 128),
+                fx=0,
+                fy=0,
+                interpolation=cv2.INTER_CUBIC)
             # HWC -> CHW
             im = im.swapaxes(1, 2)
             im = im.swapaxes(0, 1)
@@ -151,10 +149,8 @@ class FaceDetector:
 
     def Preprocess(self, image, shrink):
         h, w = int(image.shape[1] * shrink), int(image.shape[0] * shrink)
-        im = cv2.resize(image, (h, w),
-                        fx=0,
-                        fy=0,
-                        interpolation=cv2.INTER_CUBIC)
+        im = cv2.resize(
+            image, (h, w), fx=0, fy=0, interpolation=cv2.INTER_CUBIC)
         # HWC -> CHW
         im = im.swapaxes(1, 2)
         im = im.swapaxes(0, 1)
@@ -194,16 +190,18 @@ class FaceDetector:
 
 
 def predict_images(args):
-    detector = FaceDetector(model_dir=args.models_dir + '/pyramidbox_lite/',
-                            mean=[104.0, 177.0, 123.0],
-                            scale=[0.007843, 0.007843, 0.007843],
-                            use_gpu=args.use_gpu,
-                            threshold=0.7)
-
-    classifier = MaskClassifier(model_dir=args.models_dir + '/mask_detector/',
-                                mean=[0.5, 0.5, 0.5],
-                                scale=[1.0, 1.0, 1.0],
-                                use_gpu=args.use_gpu)
+    detector = FaceDetector(
+        model_dir=args.models_dir + '/pyramidbox_lite/',
+        mean=[104.0, 177.0, 123.0],
+        scale=[0.007843, 0.007843, 0.007843],
+        use_gpu=args.use_gpu,
+        threshold=0.7)
+
+    classifier = MaskClassifier(
+        model_dir=args.models_dir + '/mask_detector/',
+        mean=[0.5, 0.5, 0.5],
+        scale=[1.0, 1.0, 1.0],
+        use_gpu=args.use_gpu)
     names = []
     image_paths = []
     for name in os.listdir(args.img_paths):
@@ -229,26 +227,29 @@ def predict_video(args, im_shape=(1920, 1080), use_camera=False):
         capture = cv2.VideoCapture(0)
     else:
         capture = cv2.VideoCapture(args.video_path)
-
-    detector = FaceDetector(model_dir=args.models_dir + '/pyramidbox_lite/',
-                            mean=[104.0, 177.0, 123.0],
-                            scale=[0.007843, 0.007843, 0.007843],
-                            use_gpu=args.use_gpu,
-                            threshold=0.7)
-
-    classifier = MaskClassifier(model_dir=args.models_dir + '/mask_detector/',
-                                mean=[0.5, 0.5, 0.5],
-                                scale=[1.0, 1.0, 1.0],
-                                use_gpu=args.use_gpu)
+    detector = FaceDetector(
+        model_dir=args.models_dir + '/pyramidbox_lite/',
+        mean=[104.0, 177.0, 123.0],
+        scale=[0.007843, 0.007843, 0.007843],
+        use_gpu=args.use_gpu,
+        threshold=0.7)
+
+    classifier = MaskClassifier(
+        model_dir=args.models_dir + '/mask_detector/',
+        mean=[0.5, 0.5, 0.5],
+        scale=[1.0, 1.0, 1.0],
+        use_gpu=args.use_gpu)
 
     path = './result'
     isExists = os.path.exists(path)
     if not isExists:
         os.makedirs(path)
     fps = 30
+    width = int(capture.get(cv2.CAP_PROP_FRAME_WIDTH))
+    height = int(capture.get(cv2.CAP_PROP_FRAME_HEIGHT))
     fourcc = cv2.VideoWriter_fourcc(*'mp4v')
-    writer = cv2.VideoWriter(os.path.join(path, 'result.mp4'), fourcc, fps,
-                             args.video_size)
+    writer = cv2.VideoWriter(
+        os.path.join(path, 'result.mp4'), fourcc, fps, (width, height))
     import time
     start_time = time.time()
     index = 0
@@ -263,8 +264,12 @@ def predict_video(args, im_shape=(1920, 1080), use_camera=False):
         end_pre = time.time()
         im = VisualizeResult(frame, det_out)
         writer.write(im)
+        if args.open_imshow:
+            cv2.imshow('Mask Detection', im)
+            if cv2.waitKey(1) & 0xFF == ord('q'):
+                break
     end_time = time.time()
-    print("include read time:", (end_time - start_time) / index)
+    print("Average prediction time per frame:", (end_time - start_time) / index)
     writer.release()
 
 
@@ -273,5 +278,5 @@ if __name__ == "__main__":
     print(args.models_dir)
     if args.img_paths != '':
         predict_images(args)
-    elif args.video_path != '':
+    elif args.video_path != '' or args.use_camera:
         predict_video(args)