LoLI-Street Low-light Image Enhancement Dataset

LoLI-Street is a dataset focused on low-light image enhancement (LLIE) jointly released by research teams from Sungkyunkwan University, The Australian National University, and Tech University of Korea.LoLI-Street: Benchmarking Low-Light Image Enhancement and Beyond", and has been accepted by ACCV'24. This dataset consists of 33k pairs of low-light and well-exposed images from developed urban street scenes, covering 19k target categories for target detection. The LoLI-Street dataset also includes 1k real low-light test images for testing LLIE models under realistic conditions. It is essential for many computer vision tasks, including target detection, tracking, segmentation, and scene understanding.

The dataset was created to address the problem that existing LLIE methods perform poorly in real-world low-light conditions, especially in the street scene domain, which limits the development of robust LLIE methods. The LoLI-Street dataset helps researchers and developers train and test their models by providing images taken under a variety of low-light conditions to improve image quality and object detection in real-world applications such as autonomous driving and surveillance systems.