MTFE

Jaemin Park, An Gia Vien, Minhee Cha, Hanul Kim, and Chul Lee

Official pytorch implementation for "Multiple Transformation Function Estimation for Image Enhancement," Journal of Visual Communication and Image Representation, vol. 95, Article No. 103863, Sep. 2023.
Paper

  

Preparation

Training data: Download from GoogleDrive

The ZIP file contains three test datasets:

• LOL dataset: 485 image pairs
• FiveK dataset: 4,500 image pairs
• EUVP dataset: 11,435 image pairs

Testing samples: Download from GoogleDrive

The ZIP file contains three test datasets:

• LOL dataset: 15 image pairs
• FiveK dataset: 500 image pairs
• EUVP dataset: 515 image pairs

Pretrained weights: Download from GoogleDrive

The ZIP file contains weight files trained with each training dataset.

Training

1. Put low-quality images of training dataset in ./data/train_data/input
2. Put high-quality images of training dataset in ./data/train_data/gt
3. Put test images in ./data/test_data/LOL
4. Put ground-truths of test images in ./data/test_gt
5. Run below commend:

python lowlight_train.py

6. The trained model is saved at ./models
7. The result images are saved at ./data/analysis

Testing

1. Put test images in ./data/test_data/LOL
2. Put ground-truths of test images in ./data/test_gt
3. Run below commend:

python lowlight_test.py

4. The result images are saved at ./data/analysis

Citation

If you find this work useful for your research, please consider citing our paper:

@article{Park2023,
    author={Park, Jaemin and Vien, An Gia and Cha, Minhee and Pham, Thuy Thi and Kim, Hanul and Lee, Chul},
    booktitle={Journal of Visual Communication and Image Representation},
    title={Multiple Transformation Function Estimation for Image Enhancement}, 
    year={2023},
    volume={62},
    pages={103863},
    publisher={Elsevier}
    }
}

License

See MIT License