README.md

AIM 2020 Real Image Super-Resolution Challenge (Track1, Track2, Track3)

• paper: Wei, Pengxu, et al. "AIM 2020 challenge on real image super-resolution: Methods and results." Computer Vision–ECCV 2020 Workshops: Glasgow, UK, August 23–28, 2020, Proceedings, Part III 16. Springer International Publishing, 2020.
• Keon-Hee Ahn, MCML group, Yonsei University, South Korea (As undergraduate reaserach intern)
• email: [email protected]
• If you have any questions, please feel free to contact me through the above email

Requirements

• pytorch-gpu
• numpy
• math
• os
• argparse
• copy
• queue
• threading
• importlib
• time
• cv2
• torch.nn
• torch.optim
• torch.nn.functional

Folders

• code

  • dataloaders: folder containing dataloader
    • dataloader_multiscale.py: for loading data
  • utils: folder containing utlity
    • image_utils.py: some functions for processing data
  • model: folder which contain models
    • base.py: base for model
    • mdsr_mod5: our model
  • train_queued.py: for training models
  • get_sr.py: for acquring result images

• data: RealSR testing data for each track

  • TestX2: test for track 1
  • TestX3: test for track 2
  • TestX4: test for track 3

• experiments: trained model checkpoint for each track

  • Track1
  • Track2
  • Track3

How To Use

1. --cuda_device=0: It is device selection in case of multiple GPU. If you have only one GPU, just set it to 0.
2. --input_path, --output_path, --restore_path: you have to change 'your_directory' part in each argument to suit your environment.
3. After modifying above three arguemtns, run the following commands for each Track. Then, the result images will be created. (ex, challenge\experiments\Track1\results)

Track 1:

python get_sr.py --model=mdsr_mod5 --restore_path=your_directory\challenge\experiments\Track_1\model_200000.pth --input_path=your_directory\challenge\data\TestLRX2\TestLR --scale=2 --edsr_res_blocks=80 --output_path=your_directory\challenge\experiments\Track_1\results --cuda_device=0 --chop_forward

Track 2:

python get_sr.py --model=mdsr_mod5 --restore_path=your_directory\challenge\experiments\Track_2\model_200000.pth --input_path=your_directory\challenge\data\TestLRX3\TestLR --scale=3 --edsr_res_blocks=80 --output_path=your_directory\challenge\experiments\Track_2\results --cuda_device=0 --self_ensemble --chop_forward

Track 3:

python get_sr.py --model=mdsr_mod5 --restore_path=your_directory\challenge\experiments\Track_3\model_200000.pth --input_path=your_directory\challenge\data\TestLRX4\TestLR --scale=4 --edsr_res_blocks=80 --output_path=your_directory\challenge\experiments\Track_3\results --cuda_device=0 --self_ensemble --chop_forward