Handling Corrupted Frames in Test Video

Dear Organising Team,

I’ve encountered an issue with the test video provided with the docker. When I count the number of frames using int(cap.get(cv2.CAP_PROP_FRAME_COUNT)), it reports 62 frames, but extracting the frames only yields 60. There may be corrupted frames, which could potentially shift the frame_nr values and cause misalignment in validation.

Could you please provide guidance on how to best handle such frames? Any advice on maintaining frame consistency would be greatly appreciated.

Many thanks in advance, Kind regards, Bilal

Hi Bilal,

This is only a random test video that won't affect the final evaluation. Your container can use any method to jump frame-by-frame to generate the predictions (such as OpenCV, FFmpeg), but we recommend OpenCV since it was the package used to process and curate all datasets.

All the best, SurgVU Organizing Team