The current script for collecting images for machine learning dataset creation relies on time.sleep() to enforce a delay of 5 seconds between each image capture. This method of introducing delay not only causes the application to pause, creating a noticeable "stutter" or "freeze" in the video feed, but also makes the process of capturing screenshots less convenient and less responsive to real-time changes in the video stream. This issue of delay and unresponsiveness is evident in the instructional video, where the script's inability to capture images smoothly and promptly can lead to missed opportunities for capturing the desired frames.
Proposed Changes:
Replace the usage of time.sleep() with a non-blocking approach that utilizes cv2.VideoCapture.get(cv2.CAP_PROP_POS_MSEC) to monitor elapsed time since the video capture started. This method should allow the script to continue processing video frames while still ensuring images are captured at 5-second intervals.
Current Implementation:
for label in labels:
cap = cv2.VideoCapture(0)
print('Collecting images for {}'.format(label))
time.sleep(5)
for imgnum in range(number_imgs):
print('Collecting image {}'.format(imgnum))
ret, frame = cap.read()
imgname = os.path.join(IMAGES_PATH,label,label+'.'+'{}.jpg'.format(str(uuid.uuid1())))
cv2.imwrite(imgname, frame)
cv2.imshow('frame', frame)
time.sleep(2)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cap.release()
cv2.destroyAllWindows()
Suggested Improvement:
cap = cv2.VideoCapture(0)
for label in labels:
print('Collecting images for {}'.format(label))
initial_time = None # Initialize the start time
imgnum = 0 # Initialize the image counter
while imgnum < number_imgs:
ret, frame = cap.read()
if not ret:
print("Failed to grab frame")
break
current_time = cap.get(cv2.CAP_PROP_POS_MSEC) # Get the current timestamp of the video capture
# Initialize initial_time
if initial_time is None:
initial_time = current_time
# Save an image every 5 seconds
if current_time - initial_time >= 5000:
imgname = os.path.join(IMAGES_PATH, label, label + '.' + '{}.jpg'.format(str(uuid.uuid1())))
cv2.imwrite(imgname, frame)
print('Collecting image {}'.format(imgnum))
imgnum += 1
initial_time = current_time # Reset the start time
cv2.imshow('frame', frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cap.release()
cv2.destroyAllWindows()
Description:
The current script for collecting images for machine learning dataset creation relies on time.sleep() to enforce a delay of 5 seconds between each image capture. This method of introducing delay not only causes the application to pause, creating a noticeable "stutter" or "freeze" in the video feed, but also makes the process of capturing screenshots less convenient and less responsive to real-time changes in the video stream. This issue of delay and unresponsiveness is evident in the instructional video, where the script's inability to capture images smoothly and promptly can lead to missed opportunities for capturing the desired frames.
Proposed Changes:
Replace the usage of time.sleep() with a non-blocking approach that utilizes cv2.VideoCapture.get(cv2.CAP_PROP_POS_MSEC) to monitor elapsed time since the video capture started. This method should allow the script to continue processing video frames while still ensuring images are captured at 5-second intervals.
Current Implementation:
Suggested Improvement: