Fig. 3From: Computational imaging without a computer: seeing through random diffusers at the speed of lightGeneralization of diffractive networks that were trained with MNIST image data to reconstruct the images of different resolution test targets, seen through a known and b new randomly generated diffusers. Despite the fact that such resolution test targets or similar line-based objects were never seen by the networks during their training, their periods are successfully resolved at the output plane of the diffractive networksBack to article page