UMD Faces Facial Dataset

UMD Faces Dataset is a facial dataset mainly used for identity verification research. It has 8501 subjects and 367,920 faces. The dataset is divided into two parts: still images and video frames. The still images contain 367,888 images and 8277 subjects, while the video frames contain 22,000 subject videos and 3.7 million annotated video frames.

This dataset was released in 2016, and the main publishers are Ankan Bansal, Anirudh Nanduri, Carlos D Castillo, Rajeev Ranjan and Rama Chellappa. The related papers include "UMDFaces: An Annotated Face Dataset for Training Deep Networks; The Do's and Don'ts for CNN-based Face Verification".