# Meta-optic Accelerators for Object Classifiers

arXiv: https://arxiv.org/abs/2201.11034 · HTML: https://mlsanity.com/paper/2201.11034/

## Key facts

- Cited by 13 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 26 Jan 2022.

## Built on by (most cited first, 13 of 13)

- [Image sensing with multilayer, nonlinear optical neural networks](https://mlsanity.com/paper/2207.14293/) (2022)
- [Super-resolution image display using diffractive decoders](https://mlsanity.com/paper/2206.07281/) (2022)
- [Single-Shot Optical Neural Network](https://mlsanity.com/paper/2205.09103/) (2022)
- [Spatially Varying Nanophotonic Neural Networks](https://mlsanity.com/paper/2308.03407/) (2023)
- [Photonic Advantage of Optical Encoders](https://mlsanity.com/paper/2305.01743/) (2023)
- [Compressed Meta-Optical Encoder for Image Classification](https://mlsanity.com/paper/2406.06534/) (2024)
- [Neural Tangent Knowledge Distillation for Optical Convolutional Networks](https://mlsanity.com/paper/2508.08421/) (2025)
- [Integration of Programmable Diffraction with Digital Neural Networks](https://mlsanity.com/paper/2406.10688/) (2024)
- [Decision-making and control with diffractive optical networks](https://mlsanity.com/paper/2212.11278/) (2022)
- [Intelligent Multi-channel Meta-imagers for Accelerating Machine Vision](https://mlsanity.com/paper/2306.07365/) (2023)
- [Metasurface-generated large and arbitrary analog convolution kernels for accelerated machine vision](https://mlsanity.com/paper/2409.18614/) (2024)
- [Computational metaoptics for imaging](https://mlsanity.com/paper/2411.09133/) (2024)
- [Direct Kernel Optimization: Efficient Design for Opto-Electronic Convolutional Neural Networks](https://mlsanity.com/paper/2511.02065/) (2025)

## Abstract (arXiv)

Rapid advances in deep learning have led to paradigm shifts in a number of fields, from medical image analysis to autonomous systems. These advances, however, have resulted in digital neural networks with large computational requirements, resulting in high energy consumption and limitations in real-time decision making when computation resources are limited. Here, we demonstrate a meta-optic based neural network accelerator that can off-load computationally expensive convolution operations into high-speed and low-power optics. In this architecture, metasurfaces enable both spatial multiplexing and additional information channels, such as polarization, in object classification. End-to-end design is used to co-optimize the optical and digital systems resulting in a robust classifier that achieves 95% accurate classification of handwriting digits and 94% accuracy in classifying both the digit and its polarization state. This approach could enable compact, high-speed, and low-power image and information processing systems for a wide range of applications in machine-vision and artificial intelligence.