Open ldkong1205 opened 2 years ago
I'm not the author, but If it's still relevant I have used ffmpeg.
you need to do some preprocessing and create folders with the names of the avi files, and the while the wd is 'ucf_videos' for example, run
find ./ -mindepth 2 -name '*.avi' -execdir ffmpeg -i {} -r 1 "$filename%03d.png"
or something like that... if that doesn't work let me know
Thank you for your work.
I am running experiments on UCF -> HMDB. The dataloader in your code, however, seems need frame-level rgb images as the input instead of the original .avi video downloaded from the UCF dataset. How do I extract these rgb images from the videos downloaded directly? Could you share the code? Thank you!