The CoSal2015 co-saliency detection dataset consists of 2,015 images from 50 categories, each of which is affected by different factors, including complex environments, occlusion problems, target appearance changes, and background clutter. Cornell University released this dataset.