andreas128 / RePaint

Official PyTorch Code and Models of "RePaint: Inpainting using Denoising Diffusion Probabilistic Models", CVPR 2022
1.94k stars 157 forks source link

pretrained checkpoints of lama #24

Open Agrechka opened 1 year ago

Agrechka commented 1 year ago

Hello, I would like to know if I could use your pretrained checkpoints you used from lama. I could only find their checkpoint for celeba-hq and faces on their github. I would cite you in my current paper. Thank you.

andreas128 commented 1 year ago

Does the following work for you?

1. Download the models

pip install --upgrade gdown && bash ./download.sh

2. Apply it to the LaMa data Use the following configs to apply RePaint on LaMa thin and thick mask settings for a specific dataset.

CelebA-HQ (Faces) https://github.com/andreas128/RePaint/blob/main/confs/test_c256_thick.yml https://github.com/andreas128/RePaint/blob/main/confs/test_c256_thin.yml

ImageNet https://github.com/andreas128/RePaint/blob/main/confs/test_inet256_thick.yml https://github.com/andreas128/RePaint/blob/main/confs/test_inet256_thin.yml

Places2 https://github.com/andreas128/RePaint/blob/main/confs/test_p256_thick.yml https://github.com/andreas128/RePaint/blob/main/confs/test_p256_thin.yml

MaugrimEP commented 6 months ago

@Agrechka hey, if you retrained lama, i am interested to get the ckpt