PubFig Dataset is a large-scale face dataset, mainly used for face recognition and identification. It covers 58,797 images of 200 people on the Internet. Unlike most existing face datasets, these images are taken when the subjects are completely out of control. Therefore, there are large differences in posture, lighting, expression, scene, camera, imaging conditions and parameters in different images. This dataset is similar to the LFW dataset created by UMass-Amherst.

This dataset was released by Columbia University in 2009, and the related paper is "Attribute and Simile Classifiers for Face Verification".