[FIX] VideoManager read function failed on multiple videos

[ivan23kor]

Traceback:

"scenedetect/video_manager.py", line 757, in read
::self._downscale_factor, ::self._downscale_factor, :]
TypeError: 'NoneType' object is not subscriptable

Reason:

1. Initializing a VideoManager object with multiple video paths
2. Set the downscale factor, e.g. with set_downscale_factor()
3. Invoke read() function of the created VideoManager object, e.g. via passing it to the SceneManager detect_scenes() as frame_source argument

This happens because of the bug in video_manager.py around lines 750. If one of the captured videos is over, an empty numpy.ndarray is returned on line 745 and its further use leads to the aforementioned error.
The commit contains the bug fix in video_manager.py as well as an integration test (VideoManager + SceneManager) in tests/test_video_manager.py

[Breakthrough]

I think this should be changed to while not read_frame instead now, since if for some reason the new cap doesn't return a frame, it will still be used below. Changing this to a while() would advance to the next cap, if any.

[Breakthrough]

Sorry I realized just now, a while loop is not the solution actually - have a better suggestion.

[Breakthrough]

Hi @ivan23kor;

Great find!! I have one small favour to ask - as part of the review, I think the if statement should be changed to a while to handle the case where a video doesn't have any frames for some reason.

Do you agree/disagree? Just want to get your feedback on that before merging this.

Thank you so much for the fix, much appreciated!

[Breakthrough]

Hi @ivan23kor;

Sorry my mistake - no need for nested loops, I think it would be better to just change line 752 to continue instead. Thoughts?

Thanks!!

[ivan23kor]

The second commit is better, check it out!

[Breakthrough]

Very nice @ivan23kor! Thank you so much!!