Real-time image processing with a 2D semiconductor neural network vision sensor

Abstract

In recent years, machine vision has taken huge leaps and is now becoming an integral part of various intelligent systems, including autonomous vehicles, robotics, and many others. Usually, visual information is captured by a frame-based camera, converted into a digital format, and processed afterwards using a machine learning algorithm such as an artificial neural network (ANN). A large amount of (mostly redundant) data being passed through the entire signal chain, however, results in low frame rates and large power consumption. Various visual data preprocessing techniques have thus been developed that allow to increase the efficiency of the subsequent signal processing in an ANN. Here, we demonstrate that an image sensor itself can constitute an ANN that is able to simultaneously sense and process optical images without latency. Our device is based on a reconfigurable two-dimensional (2D) semiconductor photodiode array, with the synaptic weights of the network being stored in a continuously tunable photoresponsivity matrix. We demonstrate both supervised and unsupervised learning and successfully train the sensor to classify and encode images, that are optically projected onto the chip, with a throughput of 20 million bins per second.