DATASETS

• Please download it from here (7.1GB). Unpack the tar file to any place you want. Then, change the dir_data argument in src/option.py to the place where DIV2K images are located

• Urban100, Manga, Set5, Set15: https://cv.snu.ac.kr/research/EDSR/benchmark.tar

Running the code

python3 trainval.py -e AWSDR_SDx2_edgar -sb ./directory/to/save/logs -j 0 -r 1 # -j is for cluster, -r is for reset or start from last checkpoint

The experiment you want to run is defined with -e

Defining an experiment

Modify: exp_configs/edgar.py

Dependencies

pip install --upgrade git+https://github.com/ElementAI/haven

TODO Edgar

• Modify models/kornia_trainer.py, add the corresponding losses (commented in the code)