AFLW Face Database

AFLW Dataset is a large-scale face database with multiple poses and perspectives. Each face is annotated with 21 feature points. It contains pictures affected by factors such as different poses, expressions, lighting and race. The database has about 250 million manually annotated face pictures (59% for women and 41% for men), most of which are color pictures, and a small number of gray pictures. It is suitable for research on face recognition, face detection and face alignment, and has high research value.

This dataset was released by Graz University of Technology – Institute of Computer Graphics and Vision in 2011. The related paper is "Annotated Facial Landmarks in the Wild: A Large-scale, Real-world Database for Facial Landmark Localization".