[ECCV 2024] RRID

Image Demoireing in RAW and sRGB Domains #Paper Link

Shuning Xu, Binbin Song, Xiangyu Chen, Xina Liu and Jiantao Zhou

Updates

• ✅ 2024-03-15: Release the first version of the paper at Arxiv.
• ✅ 2024-07-01: Release the codes of RRID.
• ✅ 2024-07-11: Release the models and results of RRID.

Environment

• basicsr==1.4.2
• scikit-image==0.15.0
• deepspeed

Prepare

1. Download TMM22 dataset.
2. Download the pre-trained model.

How To Test

PYTHONPATH="./:${PYTHONPATH}" CUDA_VISIBLE_DEVICES=0 python test.py -opt options/test/Test.yml

How To Train

• Single GPU training

PYTHONPATH="./:${PYTHONPATH}" CUDA_VISIBLE_DEVICES=0 python train.py -opt options/train/Train.yml

• Distributed training

PYTHONPATH="./:${PYTHONPATH}" \
CUDA_VISIBLE_DEVICES=0,1,2,3 \
python -m torch.distributed.launch --nproc_per_node=4 --master_port=4321 train.py -opt options/train/Train.yml --launcher pytorch

Results

The inference results on benchmark datasets are available at Dropbox link.

Citations

BibTeX

@inproceedings{xu2024image,
  title={Image demoireing in raw and srgb domains},
  author={Xu, Shuning and Song, Binbin and Chen, Xiangyu and Liu, Xina and Zhou, Jiantao},
  booktitle={European conference on computer vision},
  pages={108--124},
  year={2024},
  organization={Springer}
}