This is a pytorch implementation of our research. Please refer to our paper for more details: Deep Plug-and-play Nighttime Non-blind Deblurring with Saturated Pixel Handling Schemes
- python 3.8.12 (tested with anaconda3)
- PyTorch 1.10.0 (CUDA=11.3)
pip3 install -r requirement.txt -f https://download.pytorch.org/whl/cu113/torch_stable.html
Download the DRUNet pre-trained model from DPIR
Please put the drunet_color.pth below model/pre-trained.
- The Saturated Dataset from Hu et al.
- Low-illumination Dataset from Pan et al.
- Night Dataset from Chen et al.
- Download all datasets and then put them below the Benchmark/ folder.
python3 test_deblur_benchmark.py -dataset Hu -log Log/Compare_Saturated/
python3 test_deblur_benchmark.py -dataset Pan -log Log/Compare_Low/
python3 test_deblur_benchmark.py -dataset Chen -log Log/Compare_Night/
python3 test_deblur_real_world.py
- This implementation is based on the implementation of Pytorch-template and DPIR.