Collection of 2429 constrained headshots of 277 volunteers for deep learning