The IST-EURECOM Light Field Face Database

Abstract

Light field cameras are emerging as powerful devices to capture rich scene representations that provide unique advantages for analysis and representation purposes. Some recent works have shown the power and usefulness of the richer information carried out by light field imaging, notably for face recognition. However, it is still difficult to fully assess how face recognition technology can benefit from these novel imaging sensors, notably due to the lack of appropriate test material. To support face recognition research exploiting light field images, the IST-EURECOM Light Field Face Database (IST-EURECOM LFFD) is presented in this paper. The purpose is to report the public availability of a light field face database which should be instrumental for designing, testing and validating light field imaging based recognition systems. The proposed face database includes data from 100 subjects, captured by a Lytro ILLUM camera in two 1–6 months separated sessions, with 20 samples per each person per session. To simulate multiple scenarios, the images are captured with several facial variations, covering a range of emotions, actions, poses, illuminations, and occlusions. The database includes the raw light field images, 2D rendered images and associated depth maps, along with a rich set of metadata. The IST-EURECOM LFFD is expected to become a valuable addition to existing face database repositories.