Face Reenactment

[alasokolova]

Could you please explain (or ideally upload the scripts) how to apply FlashAvatar for face reenactment task?

Should I train FlashAvatar on source head, run tracking on the target head, and then in test.py use viewpoints obtained from target and gaussians from the trained FlashAvatar?
I am a bit confused with the pipeline, so would be happy if you upload the reenactment script.

Thanks in advance!

[xiangjun-xj]

For face reenactment, you need to use relative expression transfer because the expression and shape are not well disentangled in FLAME tracking.
You can take a look at NeRFace to see how it is done.