The official implementation of paper "Multi-Outputs Is All You Need For Deblur".
-
python 3.7+
-
pytorch 1.7+
This part of code is cloned and modified from https://github.com/megvii-research/NAFNet, and some scripts are copied from https://github.com/megvii-model/HINet. For more information and guide, please see the docs in those repos.
cd NAFNet-main/
python setup.py develop --no_cuda_ext
# train
python -m torch.distributed.launch --nproc_per_node=8 --master_port=1234 basicsr/train.py -opt options/train/GoPro/MHNAFNet-width64-combinate-heads.yml --launcher pytorch
python -m torch.distributed.launch --nproc_per_node=8 --master_port=1234 basicsr/train.py -opt options/train/GoPro/MHHINet.yml --launcher pytorch
# test
python -m torch.distributed.launch --nproc_per_node=8 --master_port=1234 basicsr/test.py -opt options/test/GoPro/MHNAFNet-width64.yml --launcher pytorch
python -m torch.distributed.launch --nproc_per_node=8 --master_port=1234 basicsr/test.py -opt options/test/GoPro/MHHINet.yml --launcher pytorch
This part of code is cloned and modified from https://github.com/swz30/Restormer. For more information and guide, please see the docs in that repo.
cd Restormer-main/
python setup.py develop --no_cuda_ext
# train
./train.sh Motion_Deblurring/Options/Deblurring_MHRestormer.yml
# test
cd Motion_Deblurring
python test_multi_head.py --dataset GoPro
This part of code is cloned and modified from https://github.com/swz30/MPRNet. For more information and guide, please see the docs in that repo.
cd MPRNet-main/Deblurring
# train
python train_multi_head.py
# test
python test_multi_head.py --dataset GoPro