The CMU Pose, Illumination, and Expression (PIE) database, collected by Terence Sim, Simon Baker, and Maan Bsat, consists of over 40,000 facial images of 68 subjects. The images were captured using 13 cameras in the CMU 3D Room and a flash system similar to the Yale "fish dome," resulting in 13 poses, 43 illumination conditions, and 4 expressions. The database is designed to evaluate the robustness of face recognition algorithms to variations in pose, illumination, and expression. The paper describes the imaging hardware, collection procedure, image organization, and potential uses of the database. The PIE database is available for distribution to research groups worldwide.The CMU Pose, Illumination, and Expression (PIE) database, collected by Terence Sim, Simon Baker, and Maan Bsat, consists of over 40,000 facial images of 68 subjects. The images were captured using 13 cameras in the CMU 3D Room and a flash system similar to the Yale "fish dome," resulting in 13 poses, 43 illumination conditions, and 4 expressions. The database is designed to evaluate the robustness of face recognition algorithms to variations in pose, illumination, and expression. The paper describes the imaging hardware, collection procedure, image organization, and potential uses of the database. The PIE database is available for distribution to research groups worldwide.