pkrushith/lolv2-real

LOL v2 Dataset

Overview

The LOL v2 dataset contains 689 training images and 100 testing images for low-light image enhancement tasks. This extended version of the original LOL dataset presents more challenging scenes, providing a robust benchmark for the development and evaluation of low-light enhancement algorithms.

The data was originally downloaded from the Google Drive folder and is maintained within the Awesome-Low-Light-Enhancement GitHub repository.

Citation

If you use this dataset in your research, please cite the following paper:

Wei Chen, Wenjing Wang, Wenhan Yang, and Jiaying Liu. LOL-v2: A new benchmark for low-light image enhancement. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021.

Data Details

• Training Set: 689 images
• Testing Set: 100 images

This dataset is designed to challenge and benchmark state-of-the-art low-light image enhancement techniques with a diverse range of real-world scenarios.