sihyun-yu / PVDM

Official PyTorch implementation of Video Probabilistic Diffusion Models in Projected Latent Space (CVPR 2023).
https://sihyun.me/PVDM
MIT License
299 stars 15 forks source link

Training set used as eval/test ??? #40

Closed PallottaEnrico closed 2 months ago

PallottaEnrico commented 4 months ago

Is there a reason behind using the training set for evaluation, or is it just a mistake? https://github.com/sihyun-yu/PVDM/blob/793172f7acbe51036a9f2c62a6d2bd1b0a0c6d90/tools/dataloader.py#L305-L311

https://github.com/sihyun-yu/PVDM/blob/793172f7acbe51036a9f2c62a6d2bd1b0a0c6d90/exps/first_stage.py#L80

https://github.com/sihyun-yu/PVDM/blob/793172f7acbe51036a9f2c62a6d2bd1b0a0c6d90/exps/diffusion.py#L83

sihyun-yu commented 2 months ago

It is just for tracking the values in the evaluation. For the final evaluation in the paper we used the setup used in previous papers (StyleGAN-V).