The 20BN-something-something Dataset V2 SlowFast Trainning

# download all tar files
# extract tar files
cat 20bn-something-something-v2-?? | tar -az


# generate a video.lst with all video names(no exts) in it.
# for each video, create a path with the same name.
for video in $(cat video.lst); do mkdir ./video/${video}; done

# extract video frame
for video in $(cat video.lst); do ffmpeg -i "./tar/20bn-something-something-v2/${video}.webm" -r 30 -q:v 1 "./video/${video}/${video}_%06d.jpg"; done

## 偷懒了，训练步骤太繁琐不更新了，各位自行官方github学习
https://github.com/facebookresearch/SlowFast

## 需要数据集来研究的，可私信我
PS：我只有v2的