A group of researchers from Chinese and Canadian universities has created a dataset containing 5,000 high-resolution images of Dunhuang murals—endangered Buddhist art inside hundreds of ancient caves in Dunhuang, China—to help digitally restore and preserve them. The researchers identified 1,000 murals with obvious damages and included pixel-by-pixel annotations, which can help researchers study each mural in more granular detail and train computer vision models.
Image credit: Wikimedia Commons