.png){kind=icon}
返回
Meta-Optics Based Parallel Convolutional Processing for Neural Network Accelerator
.png){kind=icon}
OA
AI
DOI：10.1002/lpor.202300984.png)
.png){kind=icon}
.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Convolutional neural networks (CNNs) have shown great performance in computer vision tasks, from image classification to pattern recognition. However, CNNs '${\rm CNNs}<^>{\prime }$ superior performance arises at the expense of high computational costs, which restricts their employment in real-time decision-making applications. Computationally intensive convolutions can be offloaded to optical metasurfaces, enabling sub-picosecond latency and nearly zero energy consumption, but the currently reported approaches require additional bulk optics and can only process polarized light, which limits their practical usages in integrated lightweight systems. To solve these challenges, a novel design of the metasurface-based optical convolutional accelerator is experimentally demonstrated, offering an ultra-compact volume of 0.016 mm3${\rm mm}<^>{3}$, a low cross-talk of -20 dB, polarization insensitivity, and is capable of implementing multiple convolution operations and extracting simultaneously various features from light-encoded images. The ultra-compact metasurface-based optical accelerator can be compactly integrated with a digital imaging system to constitute an optical-electronic hybrid CNN, which experimentally achieves a consistent accuracy of 96 % in arbitrarily polarized MNIST digits classification. The proposed ultra-compact metasurface-based optical convolutional accelerator paves the way for power-efficient edge-computing platforms for a range of machine vision applications. This work experimentally demonstrates a novel design of the metasurface-based optical convolutional accelerator offering an ultra-compact volume of 0.016 mm3${\rm mm}<^>{3}$, low cross-talk of -20 dB, polarization insensitivity, and is capable of implementing multiple convolution operations and extracting simultaneously various features from light-encoded images. The metasurface-based optical convolutional accelerator paves the way for power-efficient edge-computing platforms for machine vision applications. image
Keyword:
convolution
machine vision
mie resonance
optical metasurfaces
photonic neural network
AI总结
对已上传原文的论文进行重点信息的提取，主要内容包括：简要概述、研究摘要、背景介绍、关键亮点、图文解析、展望与总结。
期刊
L
IF:
10
论文数:
3.7K
被引数:
2.1W
机构
引用论文
Mechanisms of 2π phase control in dielectric metasurface and transmission enhancement effect.png)
OPTICS EXPRESS
OAAI
OPTICS EXPRESSIF3.3
OAAI分享收藏Broadband and Wide Field-of-view Plasmonic Metasurface-enabled WaveplatesSCIENTIFIC REPORTSOAAI
SCIENTIFIC REPORTSIF3.9
OAAI分享收藏The cathodic behavior of microwave-treated manganese dioxide in lithium nonaqueous cells微波处理的二氧化锰在非水锂电池中的阴极行为
PREAI
分享收藏Designing of marker-based augmented reality learning environment for kids using convolutional neural network architectureDISPLAYS
DISPLAYSIF3.4
PREAI
分享收藏Flat optics with dispersion-engineered metasurfaces具有色散工程超表面的平面光学器件NATURE REVIEWS MATERIALS
NATURE REVIEWS MATERIALSIF86.2
PREAI
分享收藏
查看更多内容