search

NVlabs / few-shot-vid2vid

Pytorch implementation for few-shot photorealistic video-to-video translation.
Other
1.79k stars 276 forks source link

about pose training details #33

Closed CrossLee1 closed 4 years ago

CrossLee1 commented 4 years ago
  1. how many dancing videos collected for training pose
  2. how many frames extracted from each training video
  3. is it necessary that each video extract same number of frames

Thanks~

tcwang0509 commented 4 years ago

We collected about 1000 videos, with varying number of frames. They don't need to be the same.

Z863058 commented 4 years ago

We collected about 1000 videos, with varying number of frames. They don't need to be the same.

Great work! Can I ask What the approximate length of each video for training is? 10 seconds? 1 minutes? or longer? Thanks a lot!