search

fdbtrs / IDiff-Face

Official repository of the paper: IDiff-Face: Synthetic-based Face Recognition through Fizzy Identity-conditioned Diffusion Models (ICCV 2023)
64 stars 1 forks source link

Missing files when sampling with pre-trained model #3

Open Suimingzhe opened 11 months ago

Suimingzhe commented 11 months ago

Thanks for your job! Could you please provide the file './hydra' and 'ema_averaged_model.ckpt' in pre-trained CPD25 and CPD50 ?

NeRF-Factory commented 11 months ago

After so many hours of debugging...I finally run the code successfully. The './hydra' and 'ema_averaged_model.ckpt' are unnecessary. Instead, I modify several lines of code in 'sample. py'. If you want the code, you can add my WeChat:130008841490. I will provide a detailed explanation for you.

aksy2512 commented 9 months ago

@NeRF-Factory Can you please explain, how did u change the sampling code

NeRF-Factory commented 8 months ago

@aksy2512 Exactly, the checkpoint provided by the official repository is the weight of the 'eps_model' which is defined in the class 'DenoisingDiffusionProbabilisticModel'. You should write a new code for loading the checkpoint in the "sample.py". Additionally, the sampling process is also defined in 'DenoisingDiffusionProbabilisticModel'. You can find the code in "./diffusion/ddpm.py". This requires some modifications, but it's not too difficult.

I hope this help to answer your concern.

robbbgm commented 7 months ago

@NeRF-Factory Could you please provide us with the code?

fdbtrs commented 6 months ago

The instructions for sampling have been updated. Check the requirements list and make sure to download all required models and configuration