Fig. 7: Optoelectronic devices for in-sensor SNN.

a Greyscale images are encoded into series of spike trains by the rate coding, TTFS coding, and RTF coding schemes, and fed into a trained SNN to predict the vehicle speed and steering angle. Reproduced with permission⁷⁷. Copyright 2024, Springer Nature. b An in-sensor SNN created by combining event-driven characteristics and in-sensor computing. Reproduced with permission¹¹⁶. Copyright 2023, Springer Nature. c Output spike situation in each neuron when the three types of motions are performed sequentially. d Non-volatile and programmable photoresponsivity under a series of gate voltage pulses, proving its feasibility for use as the synaptic weights in neural networks. e Event-based sensors generate programmable spikes when light intensity changes. f Comparison between frame- and event-based vision sensors. The bottom panel shows the output signal after the pixel sensing process of event-based sensors, and only in the regions with the change of the light intensity changes generate positive or negative current spikes. Reproduced with permission⁸². Copyright 2023, Springer Nature.