Frequency-aware divide-and-conquer for efficient real noise removal (FADN)

This is the PyTorch implementation of paper: Frequency-aware divide-and-conquer for efficient real noise removal (FADN).

Diagram of FADN

pip install -r requirement.txt

Random crop SIDD for training and extract test set for testing.

python Crop_SIDD.py
python SIDD_test_extract.py

Extract DND for evaluation.

python extract_DND.py

Extract NAM for evaluation.

python extract_NAM.py

python python train.py -opt options/train/FADN.yml

Pretrained model are given in ./models/

cd codes
python test_Real.py -opt options/test/test_FADN.yml

If you have any questions, please contact huangyunlu@bupt.edu.cn.