Unconstrained face recognition from NIR sensors (often used in CCTVs) is studied substantially less than on images acquired from visible light sensors (for obvi

Unconstrained face recognition from NIR sensors (often used in CCTVs) is studied substantially less than on images acquired from visible light sensors (for obvious reasons). There's a few face recognition datasets (e.g., SCFace, MBGC Portal Challenge) that aimed to change this, but it still far from meeting the same accuracies we see from Google/Facebook.