Image from cv2.VideoCapture not working about keras-ocr HOT 4 CLOSED

It looks like you aren't checking for the video ending. So I suspect that this is working fine up until the point where the video ends, at which point ret will be False.

Immediately after ret, frame = cap.read(), you should check for the video being over by doing:

if not ret:
    break

If this does not fix it, please do the following to help with debugging the issue:

1. Change print(frame.shape) to print("Frame Shape:", frame.shape). This will make the logging easier to read.
2. Run your script and provide the 15 (or some other reasonable number) lines that are printed before the exception is raised, along with the traceback, as you have above.

from keras-ocr.

The shape is fine: (480, 640, 3) and it's the first frame of the video. Thinking about it I think I know whats wrong. I can't test it right now but I think whats happening is, that the frame has no text at all and since I use pipeline.recognize it awaits a result after the detection to feed it to the recognizer. I was about to check if text was detected only after I get predictions.
Will test this in time.

from keras-ocr.

You are 100% right. This happens whenever no text is detected in the image. I've pushed 5c1cb4f which fixes the issue and adds a test with a blank image to make sure this doesn't break again in the future. The fix is implemented in v0.6.2, which I've just published. Please try again with the updated version and let me know if the issue persists.

from keras-ocr.

Works like a charm now :)

from keras-ocr.