Does anyone know how mature algorithms that use video for facial recognition are? It seems one could use multiple consecutive frames to draw better inferences a

Does anyone know how mature algorithms that use video for facial recognition are? It seems one could use multiple consecutive frames to draw better inferences about 3D geometry, and perhaps resolve various occlusions.