Merge pull request #192 from parikshit14/master

added support for webcam

2_Training/src/keras_yolo3/yolo.py

@@ -286,3 +286,48 @@ def detect_video(yolo, video_path, output_path=""):
     vid.release()
     out.release()
     # yolo.close_session()
+
+
+# TO PROCESS VIDEOS DIRECTLY FROM WEBCAM
+def detect_webcam(yolo):
+    import cv2
+
+    vid = cv2.VideoCapture(0)
+    if not vid.isOpened():
+        raise IOError("Couldn't open webcam")
+    accum_time = 0
+    curr_fps = 0
+    fps = "FPS: ??"
+    prev_time = timer()
+    while vid.isOpened():
+        return_value, frame = vid.read()
+        if not return_value:
+            break
+        image = Image.fromarray(frame)
+        out_pred, image = yolo.detect_image(image, show_stats=False)
+        result = np.asarray(image)
+        curr_time = timer()
+        exec_time = curr_time - prev_time
+        prev_time = curr_time
+        accum_time = accum_time + exec_time
+        curr_fps = curr_fps + 1
+        if accum_time > 1:
+            accum_time = accum_time - 1
+            fps = "FPS: " + str(curr_fps)
+            curr_fps = 0
+        cv2.putText(
+            result,
+            text=fps,
+            org=(3, 15),
+            fontFace=cv2.FONT_HERSHEY_SIMPLEX,
+            fontScale=0.50,
+            color=(255, 0, 0),
+            thickness=2,
+        )
+        cv2.namedWindow("Result", cv2.WINDOW_NORMAL)
+        cv2.imshow("Result", result)
+        cv2.waitKey(1)
+        if cv2.getWindowProperty("Result", cv2.WND_PROP_VISIBLE) < 1:
+            break
+    vid.release()
+    yolo.close_session()

3_Inference/Detector.py

@@ -18,7 +18,7 @@ sys.path.append(src_path)
 sys.path.append(utils_path)
 
 import argparse
-from keras_yolo3.yolo import YOLO, detect_video
+from keras_yolo3.yolo import YOLO, detect_video, detect_webcam
 from PIL import Image
 from timeit import default_timer as timer
 from utils import load_extractor_model, load_features, parse_input, detect_object
@@ -149,12 +149,21 @@ if __name__ == "__main__":
         help="Use the tiny Yolo version for better performance and less accuracy. Default is False.",
     )
 
+    parser.add_argument(
+        "--webcam",
+        default=False,
+        action="store_true",
+        help="Use webcam for real-time detection. Default is False.",
+    )
+
     FLAGS = parser.parse_args()
 
     save_img = not FLAGS.no_save_img
 
     file_types = FLAGS.file_types
 
+    webcam_active = FLAGS.webcam
+
     if file_types:
         input_paths = GetFileList(FLAGS.input_path, endings=file_types)
     else:
@@ -215,7 +224,7 @@ if __name__ == "__main__":
     input_labels = [line.rstrip("\n") for line in class_file.readlines()]
     print("Found {} input labels: {} ...".format(len(input_labels), input_labels))
 
-    if input_image_paths:
+    if input_image_paths and not webcam_active:
         print(
             "Found {} input images: {} ...".format(
                 len(input_image_paths),
@@ -272,7 +281,8 @@ if __name__ == "__main__":
         out_df.to_csv(FLAGS.box, index=False)
 
     # This is for videos
-    if input_video_paths:
+    # for pre-recorded videos present in the Test_Images folder
+    if input_video_paths and not webcam_active:
         print(
             "Found {} input videos: {} ...".format(
                 len(input_video_paths),
@@ -293,5 +303,12 @@ if __name__ == "__main__":
                 len(input_video_paths), end - start
             )
         )
+    # for Webcam
+    if webcam_active:
+        start = timer()
+        detect_webcam(yolo)
+        end = timer()
+        print("Processed from webcam for {:.1f}sec".format(end - start))
+
     # Close the current yolo session
     yolo.close_session()

3_Inference/README.md

@@ -6,7 +6,7 @@ To detect objects run the detector script from within the [`TrainYourOwnYOLO/3_I
 ```
 python Detector.py
 ```
-The outputs are saved to [`TrainYourOwnYOLO/Data/Source_Images/Test_Image_Detection_Results`](/Data/Source_Images/Test_Image_Detection_Results). The outputs include the original images with bounding boxes and confidence scores as well as a file called [`Detection_Results.csv`](/Data/Source_Images/Test_Image_Detection_Results/Detection_Results.csv) containing the image file paths and the bounding box coordinates. For videos, the output files are videos with bounding boxes and confidence scores. To list available command line options run `python Detector.py -h`.
+The outputs are saved to [`TrainYourOwnYOLO/Data/Source_Images/Test_Image_Detection_Results`](/Data/Source_Images/Test_Image_Detection_Results). The outputs include the original images with bounding boxes and confidence scores as well as a file called [`Detection_Results.csv`](/Data/Source_Images/Test_Image_Detection_Results/Detection_Results.csv) containing the image file paths and the bounding box coordinates. For videos, the output files are videos with bounding boxes and confidence scores. For real-time detection use `python Detector.py --webcam` this will open up your webcam and detect your data classes. To list available command line options run `python Detector.py -h`.
 
 ### That's all!
 Congratulations on building your own custom YOLOv3 computer vision application.