1. To download GoPro training and testing data, run

 python download_data.py --data train-test

2. Generate image patches from full-resolution training images of GoPro dataset

python generate_patches_gopro.py 

1. Prepare you QR code and generate motion blur dataset, run

python motionblur/generate_dataset.py 

2. Generate image patches from full-resolution training images of QR code dataset

python generate_patches_qrcode.py 

1. To train EGRestormer Training(pretaining) on GoPro dataset, run

cd ADNet

torchrun --nproc_per_node=6 --nnodes=1 --node_rank=0 --master_port=4321 basicsr/train.py -opt options/train/train_egrestormer.yml --launcher pytorch

Note: The above training script uses 6 GPUs by default. To use any other number of GPUs, modify options/train/train_egrestormer.yml

2. Training (finetuning) on QRData dataset, run

torchrun --nproc_per_node=6 --nnodes=1 --node_rank=0 --master_port=4321 basicsr/train.py -opt options/train/train_egrestormer_qrdataset.yml --launcher pytorch

Note: The training iteration start from the iteration pretained on GoPro, you can modify i options/train/train_egrestormer_qrdataset.yml

1. To train LENet on Gopro, run

cd ADNet
torchrun --nproc_per_node=6 --nnodes=1 --node_rank=0 --master_port=4321 basicsr/train.py -opt options/train/train_lenet.yml --launcher pytorch

2. Training (finetuning) on QRData dataset, run

torchrun --nproc_per_node=6 --nnodes=1 --node_rank=0 --master_port=4321 basicsr/train.py -opt options/train/train_lenet.yml --launcher pytorch

Note: change the dataset in the training script.

Download the pre-trained EG-Restormer gopro_pretrain_model and place it in ./experiment/egrestormer/models/

Download the pre-trained EG-Restormer QRData_model and place it in ./experiment/egrestormer/models/

Download the pre-trained EG-Restormer QRData_model and place it in ./experiment/LENet/models/

• Download QR code test set QR code and place it in ./datasets/test/

• Testing EG-Restormer, run,

python test_egrestormer.py

• Testing LENet, run,

python test_lenet.py

• Testing ADNet, run,

python ADNet.py

If you have any question, please contact ljphit@163.com

Acknowledgment: This code is based on the Restormer and NAFNet