TCTL-Net: Template-free Color Transfer Learning for Self-Attention Driven Underwater Image Enhancement
This repository is the official PyTorch implementation of TCTL-Net.
- For training, you need to prepare the raw, ref, and txt folders in the train data root directory to store the raw degraded image, the reference enhanced image, and the reference true value of color-transfer, respectively.
- For testing, you need to prepare the raw folder in the test data root directory to store the test images.
python train.py --dataroot /path_to_data --name train_name
python test.py --dataroot /path_to_data --name test_name
You can download the trained model from Baidu Netdisk(Extraction code:rw85) or Google Drive.