search

jingyuanli001 / RFR-Inpainting

The source code for CVPR 2020 accepted paper "Recurrent Feature Reasoning for Image Inpainting"
MIT License
355 stars 76 forks source link

Training on Celeba-HQ datasets #28

Closed laolongboy closed 2 years ago

laolongboy commented 3 years ago

image image I am not sure the results are reliable. Will u release Celeba-hq models?

jingyuanli001 commented 3 years ago

Hi. Unfortunately, we are not planning to release the model for Celeb-HQ. However, when we were developing the model, we've indeed trained our model on CelebA-HQ dataset and the results are also satisfactory. For your case, I would suggests you to train the model for a longer time, and be sure to do adequate finetuning before testing (which is significant for stabilize the performance). You might also insert one or two extra partial convolution layers to down-sample the images in the inpainting procedure.