Confused about the strategy of Mixed image and video training in paper.

KwaiVGI / LivePortrait

Bring portraits to life!

https://liveportrait.github.io

Other

12.54k stars 1.32k forks source link

Confused about the strategy of Mixed image and video training in paper. #208

Open XiaoluDing opened 3 months ago

XiaoluDing commented 3 months ago

Amazing work，what about 'To leverage both data types, we treat single images as one-frame video clips and train the model on both images and videos.'

when treat a single image as a clip in training stage 1 ，How to differentiate between source and driver images，Or does the 60K images not learn the differences in stage1?

Vincent-luo commented 2 weeks ago

@XiaoluDing Have you figured it out? I'm also a little confused.