Face-in-Action Dataset is a facial video dataset with imaging variations. It covers 20 seconds of facial video data from 180 participants. It is mainly used to imitate passport inspection scenarios. The relevant videos are captured by 6 synchronized cameras at 3 different angles (the focal length of each angle is 8mm and 4mm). The images are collected in restricted indoor environments and open outdoor environments.

This dataset was released by Carnegie Mellon University in 2010.