10,000 Cat Pictures (For Science)
Researchers from Microsoft Research Asia and the Chinese University of Hong Kong created a database with 10,000 photos containing cat heads to test image recognition algorithms, and now the data set is available freely for research purposes. The photos, which the researchers downloaded mainly from Flickr, are paired with data files that specify the location of each cat’s eyes, mouth, and ears. Using these features with their algorithmic method, the researchers were able to distinguish a cat photo from a non-cat photo with better accuracy than previous methods. The creators of the collection, which is affectionately known as the CAT Database, hope their data might be useful to other machine learning researchers working in facial recognition, image comprehension, and other computer vision topics.