Fictionarry / ER-NeRF

[ICCV'23] Efficient Region-Aware Neural Radiance Fields for High-Fidelity Talking Portrait Synthesis
https://fictionarry.github.io/ER-NeRF/
MIT License
894 stars 124 forks source link

3DMM #90

Open deligenters opened 7 months ago

deligenters commented 7 months ago

What is the function of the 3DMM module?

husthzy commented 7 months ago

According to the parameters required for NeRF training, the 3DMM face tracker is used to obtain the internal and external camera parameters corresponding to each frame.

xuxianren commented 5 months ago

According to the parameters required for NeRF training, the 3DMM face tracker is used to obtain the internal and external camera parameters corresponding to each frame.

If the video is shot with a fixed camera position, will the camera parameters calculated for each frame be the same?

husthzy commented 5 months ago

According to the parameters required for NeRF training, the 3DMM face tracker is used to obtain the internal and external camera parameters corresponding to each frame.

If the video is shot with a fixed camera position, will the camera parameters calculated for each frame be the same?

The internal parameters of the camera are fixed. The external parameter matrix is a matrix composed of the 3D position and posture angle of the face in the camera coordinate system obtained in each frame. Since the human face is moving, the extrinsic parameter matrices of each frame are different.