A volume holographic film (such as a photopolymer) that is pre-recorded with patterns subsequently is used to encode LED or low-power laser light reflections from an eye into a binary pattern that can be read at very high speeds by a relatively simple complementary metal-oxide-semiconductor (CMOS) sensor that may be similar to a high framerate, low resolution mouse sensor. The low-resolution mono images from the film are translated into eye poses using, for instance, a look up table that correlates binary patterns to X, Y positions or using a pre-trained convolutional neural network to robustly interpret many variations of the binary patterns for conversion to X, Y positions.
