Apple has filed a patent application for an improved version of Face ID that would be able to recognize partial faces and overcome poor lighting conditions. The newly published application was filed with the U.S. Patent and Trademark Office.
The patent specifically concerns a neural network module for a device equipped with a TrueDepth camera. Whenever it detects a face in an image, the network will be able to create a bounding box and analyze that face regardless of its orientation within the frame. The box will be created as long as the face is at least partially visible within the image.
Apple did not offer as many details about how it plans to deal with adverse lighting, and could reveal more information in a separate patent application. Facial recognition providers like Aratek and Panasonic have previously boasted about the low lighting capabilities of their technology.
The new patent could facilitate the integration of Face ID into a slew of IoT devices, especially if those devices capture faces at unusual angles. For example, a recent Apple patent detailed plans to introduce Face ID to the Apple Watch, which could otherwise prove to be more challenging than it is with a standard iPhone.
Source: Patently Apple