Related Experiment Video
Updated: Oct 14, 2025
06:25
Author Spotlight: Assessment of Visual Acuity in Central Vision Loss Through Motion-Based Peripheral Vision Testing
Published on: February 23, 2024
761
All-in-one two-dimensional retinomorphic hardware device for motion detection and recognition.
Zhenhan Zhang1,2, Shuiyuan Wang1, Chunsen Liu1,3
1State Key Laboratory of ASIC and Systems, School of Microelectronics, Fudan University, Shanghai, China.
Nature Nanotechnology
|November 9, 2021
Summary
Researchers developed a novel 2D retinomorphic device for efficient motion detection and recognition (MDR). This retina-inspired hardware integrates perception, memory, and computing, enabling compact and high-performance object tracking.
Area of Science:
- Materials Science
- Computer Engineering
- Neuroscience
Background:
- Current motion detection and recognition (MDR) systems using CMOS image sensors (CIS) are inefficient due to redundant modules.
- Existing non-memory vision sensors are limited to static target detection, not dynamic object recognition.
- The human retina offers a highly efficient model for integrated visual perception, memory, and computation.
Purpose of the Study:
- To develop a compact and efficient hardware device for motion detection and recognition (MDR).
- To create a novel, retina-inspired 2D heterostructure device with integrated perception, memory, and computing capabilities.
- To demonstrate the device's effectiveness in detecting and recognizing moving objects, specifically trolleys.
Main Methods:
- Fabrication of a two-dimensional (2D) heterostructure based retinomorphic hardware device.
- Utilizing the device's ability to generate tuneable photoresponses and memorize optical stimuli.
- Implementing interframe differencing computations and a conductance-mapped neural network for recognition.
Main Results:
- Achieved 100% separation detection of moving trichromatic trolleys without ghosting.
- Demonstrated fast trolley recognition using a conductance-mapped neural network in as few as four training epochs.
- Showcased superior performance compared to previous results on similar datasets, even at a 10% noise level.
Conclusions:
- The proposed 2D retinomorphic device successfully integrates perception, memory, and computation for effective MDR.
- This prototype demonstrates the potential for creating compact, efficient, and high-performance motion detection and recognition hardware.
- The retina-inspired design offers a promising pathway for next-generation vision sensing technologies.
