Is your feature request related to a problem? If so, Please describe.
Currently, we don't have anything in place to support training on videos, we assume the input data is an image sequence.
Describe the solution you'd like
A simple solution would be to use ffmpeg to dump the video into an image sequence, and then train on that. Suggested by @mattdawkins.
Describe alternatives you've considered
Another solution is for @mattdawkins to add video support to the viame training tool itself. Then, we would just pass in the unmodified video.
I'm not sure which of these approaches is better, and or what the pros/cons are in either case. I think ideally handling video in the training tool itself would be preferable, but @mattdawkins mentioned that the functionality to support that might just involve dumping the video to an image-sequence in the tool anyway.
Is your feature request related to a problem? If so, Please describe.
Currently, we don't have anything in place to support training on videos, we assume the input data is an image sequence.
Describe the solution you'd like
A simple solution would be to use
ffmpegto dump the video into an image sequence, and then train on that. Suggested by @mattdawkins.Describe alternatives you've considered
Another solution is for @mattdawkins to add video support to the viame training tool itself. Then, we would just pass in the unmodified video.
I'm not sure which of these approaches is better, and or what the pros/cons are in either case. I think ideally handling video in the training tool itself would be preferable, but @mattdawkins mentioned that the functionality to support that might just involve dumping the video to an image-sequence in the tool anyway.