Progress in Bioinspired Photodetectors Design for Visual Information Processing

Abstract

Rapid development of modern science and technology has prompted explosive growth of data volumes, especially the visual information, which brings heavy pressure on information processing and computation. The classic technical route to improve the computation power of image processing units by reducing the critical dimension of the transistors according to Moore's law gradually fails due to the physical limit of transistors’ footprint and the separated architectures. Inspired by human vision systems, retina‐like photodetectors that mimic their spatial and temporal properties have attracted widespread attention. These developed architectures enable the early data processing in‐ or near‐sensor, significantly reducing the computing power required in the back end. Herein, a comprehensive review on bioinspired photodetectors that possess the spatial and temporal properties of biological vision is provided. The properties of retina are summarized and presented first. Basic structure design and operation mechanism of those photodetectors with spatial properties of retina (including the distribution of receptive fields and the shape of the hemisphere) and temporal properties of retina (including the memory properties enabled learning and the light intensity adaptation) are reviewed thoroughly. Eventually, challenges and future perspectives are commented and provided to facilitate the rapid development of in‐ or near‐sensor computing photodetectors.

A Reconfigurable All-Optical-Controlled Synaptic Device for Neuromorphic Computing Applications

“…The fundamental component of neuromorphic visual sensors is an optoelectronic artificial synaptic device that simulates synaptic activity in response to electrical or optical stimuli. In complex neural networks, synapses are the basic units of computation and learning, acting as connections between neurons. ,, It is believed that synaptic plasticity, or variations in synaptic weight, under different stimulus conditions, is thought to underlie the logic, learning and memory processes in the brain. , In general, synapses can respond to various external stimuli with excitatory and inhibitory synaptic activity, and exhibit short-term plasticity or long-term plasticity depending on the duration of decay following the cessation of external spikes. ,, Hence, the ability to accurately replicate synaptic plasticity under light stimulation is crucial for the advancement of optoelectronic neuromorphic devices. Nevertheless, due to the fact that optical artificial synapses exhibit a unidirectional response to light stimuli, both optical and electrical coconfigurations are necessary to accomplish reversible modulation of synaptic weights in most of the reported photoelectric synapses. ,,− Moreover, optical stimulation often relates to the potentiation process, whereas electrical stimulation acts on the depression process.…”

“…22,23 In general, synapses can respond to various external stimuli with excitatory and inhibitory synaptic activity, and exhibit short-term plasticity or long-term plasticity depending on the duration of decay following the cessation of external spikes. 5,20,24 Hence, the ability to accurately replicate synaptic plasticity under light stimulation is crucial for the advancement of optoelectronic neuromorphic devices. Nevertheless, due to the fact that optical artificial synapses exhibit a unidirectional response to light stimuli, both optical and electrical coconfigurations are necessary to accomplish reversible modulation of synaptic weights in most of the reported photoelectric synapses.…”

“…Inspired by biological vision, neuromorphic-based visual sensors have shown great advantages and received increasing attention for realizing image preprocessing and intelligent detection. − The von Neumann architecture is traditionally employed in machine vision systems. , The sensor transforms a substantial quantity of visual data into a digital electronic format in a storage unit, which is then passed to a computing unit for processing. Data movement occurs as a consequence of computation delays, excessive energy consumption, and data redundancy due to the physical separation between the storage unit and the computing unit. − Nevertheless, in the age of artificial intelligence, there is a growing need for extensive visual data processing and real-time detection.…”

Engineering Graphene Phototransistors for High Dynamic Range Applications

“…The advancement of low-dimensional materials has spurred the development of various optical applications including biosensors, bioimaging, and photodetectors. − The advancements in photodetectors, among others, are driven by the desire to reproduce and improve the capabilities of the human sensory system. Such photodetection devices have moved beyond their traditional role of linearly measuring light intensity and are implemented in a wide range of applications, from neural networks and neuromorphic photodetection , to artificial vision systems , and vision restoration. , In the context of vision-related applications, it becomes imperative to consider the characteristics of the human visual system. Throughout the transition from dawn to dusk, typical retinal illuminance levels encompass a range from 1 to 10 5 lx .…”

“… 3 − 5 The advancements in photodetectors, among others, are driven by the desire to reproduce and improve the capabilities of the human sensory system. Such photodetection devices have moved beyond their traditional role of linearly measuring light intensity and are implemented in a wide range of applications, from neural networks and neuromorphic photodetection 6 , 7 to artificial vision systems 8 , 9 and vision restoration. 10 , 11 In the context of vision-related applications, it becomes imperative to consider the characteristics of the human visual system.…”