Guy Gafni
Hey, if you use the expressions and poses from one actor with a trained model of another, you will basically get reenactment. There are a few things to be careful...
Since nerf requires static object at a canonical space and cameras revolving around it, we use the transformation matrix of the rigid head tracking as the camera rigid transformations. (And...
You need a face tracker, you can try some open source ones, eg [[](url)](https://github.com/philgras/video-head-tracker)
Line 420 in the eval script is causing the pose and expression to be fixed. Set it to None
Sorry for leaving all the last minute ablation mess there. You were right to remove/comment out/None anything related to ablation.
Tips for cross subject reenactment would be to select a neutral expression looking frame from each, and then only transfer the deltas from that. This reduces the domain gap. if...
If you use COLMAP to solve for the camera poses, you can use the LLFF dataloader after reading the COLMAP output. You can find more details about this in the...
@yangqing-yq yes, you can just change the 76 to the dimension of the FLAME expression vector. With the rigid pose you have to be a bit more careful, as FLAME...
About 48h on a low end GPU (2080). Maybe you're running on CPU, or have wrong hyperparams
Hi, this is just a mapping between temporal index of a frame and it's index in the training set. I used it to retrieve latent codes of specific timestamps, or...