Added Webcam.

README.md

@@ -19,6 +19,15 @@ Original paper and github: [YOLO9000: Better, Faster, Stronger](https://arxiv.or
 - [opencv](https://pypi.org/project/opencv-python/)
 - [python3.5](https://www.python.org/)
 
+### Simple use
+---
+1. Download pretrained [model](https://drive.google.com/open?id=1khOgS8VD-paUD8KhjOLOzEXkzaXNAAMq) and place it in **dourflow/**.
+2. Predict on an image:
+
+```bash
+python3 dourflow.py path/to/test_image
+```
+
 ### Usage
 ---
 Running `python3 dourflow.py --help`:
@@ -45,8 +54,9 @@ optional arguments:
 Pass what to do with dourflow:
 
 1. A path to an image file/dir or video: Run inference on those file(s).
-2. 'validate': Perform validation on a trained model.
-3. 'train': Perform training on your own dataset.
+2. 'cam': Run inference on webcam ('cams' to store the output).
+3. 'validate': Perform validation on a trained model.
+4. 'train': Perform training on your own dataset.
 
 ##### *model*
 Pass the keras input model h5 file (could be to perform inference, validate against or for transfer learning). 
@@ -128,10 +138,12 @@ python3 dourflow.py images/ -m coco_model.h5 -c coco_config.json -t 0.35
 Allows to evaluate the performance of a model by computing its [mean Average Precision](http://host.robots.ox.ac.uk:8080/pascal/VOC/voc2012/htmldoc/devkit_doc.html#SECTION00050000000000000000) in the task of object detection (mAP WRITE UP COMING SOON).
 
 Example:
+
 ```bash
 python3 dourflow.py validate -m voc_model.h5 -c voc_config.json
 ```
 Output:
+
 ```bash
 Batch Processed: 100%|████████████████████████████████████████████| 4282/4282 [01:53<00:00, 37.84it/s]
 AP( bus ): 0.806
@@ -162,10 +174,6 @@ mAP: 0.674
 
 ### Training
 ---
-**NEED TO DO A FEW FIXES TO LOSS FUNCTION BEFORE THIS IS DONE**
----
-
-
 
 ##### Split dataset
 Script to generate training/testing splits.
@@ -193,11 +201,10 @@ Then, in another terminal tab you can run `tensorboard --logdir=logs/run_X` and
 
 #### To Do
 
-- [ ] TRAINING BUG / FINISH LOSS FUNCTION
 - [ ] cfg parser
 - [ ] Anchor generation for custom datasets
 - [ ] mAP write up
-- [ ] Add webcam support
+- [x] Add webcam support
 
 #### Inspired from
 

net/netarch.py

@@ -30,7 +30,7 @@ def get_model(self, loss_func):
             #new_name = self.tl_weights_name.split('.')[0] + '_rand.h5'
             #new_yolo_model.save_weights(new_name)
 
-        elif YoloParams.YOLO_MODE in ['inference','validate','video']:
+        elif YoloParams.YOLO_MODE in ['inference','validate','video','cam']:
             new_yolo_model = yolo_model
 
         else:

net/netparams.py

@@ -17,14 +17,14 @@
     '-m',
     '--model',
     help='path to input yolo v2 keras model',
-    default='yolo_model.h5')
+    default='coco_model.h5')
 
 
 argparser.add_argument(
     '-c',
     '--conf',
     help='path to configuration file',
-    default='config.json')
+    default='confs/config_coco.json')
 
 
 argparser.add_argument(
@@ -71,8 +71,16 @@ class YoloParams(object):
     # Mode
     PREDICT_IMAGE = ''
     WEIGHT_FILE = ''
-    if action != 'gen':
-        if action == 'validate' or action == 'train':
+    WEBCAM_OUT = ''
+
+    if action == 'gen':
+        assert args.weight_file, "Need to pass weight file if generating model."
+        WEIGHT_FILE = args.weight_file
+    elif action == 'cams':
+        WEBCAM_OUT = 'cam_out.mp4'
+        YOLO_MODE = 'cam'
+    else:
+        if action == 'validate' or action == 'train' or action == 'cam':
             YOLO_MODE = action
         else:
             if os.path.isdir(action):
@@ -83,14 +91,10 @@ class YoloParams(object):
                 else:
                     YOLO_MODE = 'inference'
             else:
-                raise ValueError('First argument for dourflow must be: \'training\','
-                    ' \'validation\' or an image file/dir.')    
+                raise ValueError('Run \'python3 dourflow.py --help\'.')    
 
             PREDICT_IMAGE = action
-    else:
-        assert args.weight_file, "Need to pass weight file if generating model."
-        # Paths
-        WEIGHT_FILE = args.weight_file
+
 
     TRAIN_IMG_PATH = config['train']['image_folder'] 
     TRAIN_ANN_PATH = config['train']['annot_folder']

yolov2.py

@@ -91,6 +91,10 @@ def run(self, **kwargs):
             elif YoloParams.YOLO_MODE == 'video':
                 self.video_inference(YoloParams.PREDICT_IMAGE)
 
+            elif YoloParams.YOLO_MODE == 'cam':
+                self.cam_inference(YoloParams.WEBCAM_OUT)
+
+
         # Sometimes bug: https://github.com/tensorflow/tensorflow/issues/3388
         K.clear_session()
 
@@ -136,8 +140,7 @@ def video_inference(self, filename):
         size = (video_width, video_height)
         fps = round(cap.get(cv2.CAP_PROP_FPS))
 
-        #fourcc = cv2.VideoWriter_fourcc(*"MJPG")
-        fourcc = cv2.VideoWriter_fourcc(*"XVID")
+        fourcc = cv2.VideoWriter_fourcc('m','p','4','v')
 
         writer = cv2.VideoWriter(filename.split('.')[0]+"_pred.mp4", fourcc, fps, size)
 
@@ -164,31 +167,36 @@ def video_inference(self, filename):
         cv2.destroyAllWindows()
 
 
-    def video_inference2(self):
-        # MAKEONE FOR WEBCAM, USE COMMENTED CODE FROM ABOVE
-        video_reader = cv2.VideoCapture(video_inp)
+    def cam_inference(self, fname):
+
+        cap = cv2.VideoCapture(0)    
+        video_width = int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))
+        video_height = int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))
+        video_len = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
+        size = (video_width, video_height)
+        fps = round(cap.get(cv2.CAP_PROP_FPS))
 
-        nb_frames = int(video_reader.get(cv2.CAP_PROP_FRAME_COUNT))
-        frame_h = int(video_reader.get(cv2.CAP_PROP_FRAME_HEIGHT))
-        frame_w = int(video_reader.get(cv2.CAP_PROP_FRAME_WIDTH))
+        fourcc = cv2.VideoWriter_fourcc('m','p','4','v')
 
-        video_writer = cv2.VideoWriter(video_out,
-                                       cv2.VideoWriter_fourcc(*'XVID'), 
-                                       50.0, 
-                                       (frame_w, frame_h))
+        if fname: writer = cv2.VideoWriter("out.mp4", fourcc, fps, size)
 
+        while(cap.isOpened()):
 
-        for i in tqdm(range(nb_frames)):
-            ret, image = video_reader.read()
+            ret, frame = cap.read()
+            if ret==True:
 
-            boxes, scores, _, labels = self.inf_model.predict(image)
+                boxes, scores, _, labels = self.inf_model.predict(frame)
+                frame_pred = draw_boxes(frame, (boxes, scores, labels))
 
-            image = draw_boxes(image, (boxes, scores, labels))
-
-            video_writer.write(np.uint8(image))
-
-        video_reader.release()
-        video_writer.release()  
+                if fname: writer.write(frame)
+
+                cv2.imshow('Yolo Output',frame)
+                if cv2.waitKey(1) & 0xFF == ord('q'):
+                    break
+
+        cap.release()
+        if fname: writer.release()
+        cv2.destroyAllWindows()