FeiiYin

Sorry for the mistake, I release a cleaner requirements.txt for reference.

I cannot load the videos correctly to see the results. But indeed, the GAN inversion method is not perfect and the warping procedure will raise artifacts in the cross-id scenarios....

Thanks for your interest in our project. We will release the code by June at the latest upon approval.

Sorry, our paper has not been accepted yet.

The class NoVGGHPMCorrespondence class requires the input of segmentation. What is the segmantic infomation of sketches or edges? How to achieve the training process if my data do not own...

That is weird since there is no error reported. Maybe it is processing the 3DMM extraction. You can print log in inner modules to check which code line it stuck...

Sorry for the mistake, `train.py` file has been released and the training configs in yaml have been fixed.

Since our model is finally trained on the HDTF dataset in an end-to-end way, the performance is quite related to the distribution of the dataset. The roll angle in HDTF...

Sorry, we are currently not able to share the audio part checkpoint.

Ok, now I can watch the videos. I recommend using the optimization-based inversion methods, e.g. PTI or BaberShop to inverse the target identity. The identity will be strongly improved compared...