search

zhengkw18 / face-vid2vid

Unofficial implementation of the paper "One-Shot Free-View Neural Talking-Head Synthesis for Video Conferencing" (CVPR 2021 Oral)
153 stars 17 forks source link

about the nework #4

Open marvin-nj opened 2 years ago

marvin-nj commented 2 years ago

Hello, zhengkw18, thank you for your contribution!

the output “delta” of the the HPE_EDE model should be the expression of the persion, not head pose, right ? but , when i frozen the yaw,pitch and roll matrixs, and only extract delta feature from HPE model of driving person , the source persion still have a head movtion. so , what's wrong with me?

I want to transfer one person's expression from another, with no head movtion. how shoud i do.

zhengkw18 commented 2 years ago

My pretrained model is insufficiently trained so the decorrelation between the components is not guaranteed.

marvin-nj commented 2 years ago

thanks! so , the expression feature can be decomposed if model trained sufficiently， right ?

zhengkw18 commented 2 years ago

Theoretically it is. You can see the NVlab's demo, which demonstrates quite impressive results.