Dependency

• Python 3.6
• PyTorch
• glob
• numpy
• pillow
• tqdm
• tensorboardx
• PIL
• basicsr

Code

• Train_SPLUT_L/M/S.py: the code for training SPLUT
• Transfer_SPLUT_L/M/S.py: the code for transferring mapping modules to LUTs
• Inference_SPLUT_L/M/S.py: the code for testing SPLUT

Training

1. Prepare folders of ./log and ./transfer
2. Prepare DIV2K training images in ./train

• HR images: ./train/DIV2K_train_HR/*.png
• LR images:./train/DIV2K_train_LR_bicubic/X4/*.png

3. Prepare LR/HR images in ./val

4. Run python Train_SPLUT_L/M/S.py

5. Checkpoints used in our paper have been saved in ./checkpoint

• Training log will be generated in ./log.

Transferring

1. Run python Transfer_SPLUT_L/M/S.py

2. The resulting LUT will be saved in the folder of ./transfer/

Inference using LUT

1. The Set5 LR/HR images for testing are already included in ./test_dataset. The users can also evaluate other images.
2. Run python Inference_SPLUT_L/M/S.py