TBSN: Transformer-Based Blind-Spot Network for Self-Supervised Image Denoising Arxiv
Download SIDD and DND datasets, and modify dataset_path
in dataset/base_function.py
accordingly.
|- dataset_path
|- SIDD
|- SIDD_Medium_Srgb
|- Data
|- 0001_001_S6_00100_00060_3200_L
|- 0002_001_S6_00100_00020_3200_N
|- ...
|- SIDD_Validation
|- ValidationNoisyBlocksSrgb.mat
|- ValidationGtBlocksSrgb.mat
|- SIDD_Benchmark
|- BenchmarkNoisyBlocksSrgb.mat
|- DND
|- info.mat
|- images_srgb
Validate on SIDD Validation dataset,
cd validate
python base.py --config_file "../option/tbsn_sidd.json"
Training on SIDD Medium dataset,
sh train.sh