EverybodyDanceNow_reproduce_pytorch icon indicating copy to clipboard operation
EverybodyDanceNow_reproduce_pytorch copied to clipboard

Published 20 hours ago • verified publisher icon yanx27

about time-smoothing

Open zrhyst23 opened this issue 4 years ago • 1 comments

The time smoothing module mentioned in the paper means that during training, the input of the generator is the 2D posture of the human body(xt) and the generated picture of the previous frame（G（xt-1））. so is it the same during the test? When testing, do we need to put the 2D pose of the human body (xt)and the generated picture of the previous frameG（xt-1）） every time? I know you didn’t do this part, but I just want to konw what do you think about it. Thanks a lot!

zrhyst23 avatar Jun 11 '20 09:06 zrhyst23

Hi, I also want to add the time smoothing module into this implementation, but I just have no idea what to do with the training period. Could you please give me some advice or guidance?
Additionally, I don't think we need to introduce inter-frame correlation during the test, cause the model is already trained. Thanks!

tyrink avatar Dec 16 '20 13:12 tyrink