Official Code for ICCV 2021 paper "Towards Flexible Blind JPEG Artifacts Removal (FBCNN)"
Pre-trained models of FBCNN:
* fbcnn_color.pth # trained with single jpeg color images
* fbcnn_gray.pth # trained with single jpeg grayscale images
* fbcnn_gray_double.pth # trained with double jpeg grayscale images (FBCNN-A in our paper)