Dark_Night

We present a real-world nighttime dataset designed for research in low-light image enhancement and night-time computer vision tasks. The dataset consists exclusively of low-light images captured under diverse nighttime conditions, including urban streets, highways, and indoor environments. Images are unpaired, making the dataset particularly suitable for unsupervised or self-supervised learning techniques. Due to its unpaired nature, the dataset is ideal for evaluation using no-reference perceptual quality metrics such as NIQE, BRISQUE, and PIQE. This dataset provides a challenging benchmark for developing robust algorithms applicable to night-time perception, autonomous driving, and mixed-reality applications.