search

Qiulin-W / SAFA

Official Pytorch Implementation of 3DV2021 paper: SAFA: Structure Aware Face Animation.
Other
177 stars 29 forks source link

Please advise on how to increase resolution #9

Closed VisionaryMind closed 2 years ago

VisionaryMind commented 2 years ago

First, I want to congratulate you on an ingenious paper. The idea of using FLAME with face alignment and masking is simply brilliant. I am getting very good results with your current code. However, I would like to ask if you could provide some brief advice on how to raise the resolution (e.g. from 256x256 to 512x512). I attempted to adjust settings in the code, but there were persistent tensor mismatch problems. I presume the dataset must be re-trained at higher resolution? Please advise, if you will have time.

Qiulin-W commented 2 years ago

Hi, thank for your interests to our paper. Regarding to increasing resolution, you need to change the resolution of the rendered 3D reenactment and 3D face motion so that they can fit.

VisionaryMind commented 2 years ago

@Qiulin-W, thank you for the explanation. This will require a re-training of the model using the VoxCeleb dataset. We have tried to download it using the Youtube URLs, but even with 32 workers, this will be a ~20 day task. Would you be so kind as to provide the dataset (temporarily) for download? The re-training is expected to take much time, and having a faster way to procure the data would speed up the process. Thank you!

Qiulin-W commented 2 years ago

Sorry we can't provide the dataset.