"BALit: Dataset for Backlit Low-Light Image Enhancement"

"BALit is a paired image dataset designed for backlit low-light image enhancement. It contains 1,000 real-world backlit low-light RAW images captured mainly under outdoor transitional lighting conditions, together with expert-retouched reference images prepared using Adobe Camera Raw. The dataset focuses on scenes where global underexposure and localized backlighting appear together, producing dark noisy shadows, strong highlight regions, and large dynamic-range imbalance within the same image. BALit includes diverse scene categories such as people, architecture, natural scenes, objects, streets, vehicles, and urban assets. The original inputs are preserved as Canon RAW files, while the references are provided as high-quality image files for supervised enhancement research. The dataset is intended to support research on low-light enhancement, backlit image enhancement, RAW-to-RGB restoration, tone correction, shadow recovery, highlight preservation, and robust multimedia image processing."