Artificial intelligence (AI) has been used to design 3D-printed lenses that can filter out extraneous objects so that only the target object appears in a photograph.
Deep learning was used by the system to tweak a lens design to pass certain objects (numbers, trainers or trousers) as target objects to a camera sensor for imaging whilst filtering out other objects in the field of view.
Interestingly, all of the processing was achieved by the physical lens design, which diffracted light from non-target objects so that it never reached the camera sensor. Accordingly, no computer-based image editing was required. The downside being that each lens is only suitable for a specific target object or shape, with different lenses being required to filter for different objects.