CoPhIR Dataset is a dataset consisting of about 106 million images collected from Flickr, which is mainly used for image classification. The images not only contain the data of the chart itself, such as location, title, tags, comments, etc., but also can extract color patterns, color layout, edge histograms, uniform textures, etc.

This dataset was released by the NMIS laboratory in 2009. Related papers include "CoPhIR: a Test Collection for Content-Based Image Retrieval" and "Enabling Content-Based Image Retrieval in Very Large Digital Libraries".