search

zhanglonghao1992 / One-Shot_Free-View_Neural_Talking_Head_Synthesis

Pytorch implementation of paper "One-Shot Free-View Neural Talking-Head Synthesis for Video Conferencing"
Other
735 stars 141 forks source link

Visulization Problem during Training #47

Closed Amazingren closed 2 years ago

Amazingren commented 2 years ago

Hi @zhanglonghao1992,

Thanks for your inpressive contribution to this work. I want to retrain by myself first on this project, and I followed every step in this repo and fomm. However, when I successfully run the training code after nearly 3 hours on 2gpus. I found the visulization is just as follows: image

Only the firsrt several colums showed some figures, hence I am afriad that there maybe some problem for my training. Do you have any idea on this kind of problem? Please let me know if there are any suggestion~

Thanks very much and best regards BTY, my environments:

Amazingren commented 2 years ago

I got it.

It is not a problem actually.

We need patient to wait the results. When my experiments went to epoch 6, it then shows like this: image