Machine‐Learned Light‐Field Camera that Reads Facial Expression from High‐Contrast and Illumination Invariant 3D Facial Images