Related Experiment Video
Updated: Apr 22, 2026
11:24
Targeted Labeling of Neurons in a Specific Functional Micro-domain of the Neocortex by Combining Intrinsic Signal and Two-photon Imaging
Published on: December 12, 2012
12.9K
Metasurface-Based Wavelength-Multiplexed Diffractive Neural Networks for Multilabel Intelligent Vision.
Rui Yang1,2, Lei Chen1,2, Zhao Wang1,2
1School of Artificial Intelligence Science and Technology, University of Shanghai for Science and Technology, Shanghai 200093, China.
Nano Letters
|April 21, 2026
Summary
This study introduces a wavelength-multiplexed diffractive neural network (WMDNN) for advanced optical AI. The WMDNN enables simultaneous object and color recognition, overcoming limitations of single-label systems.
Area of Science:
- Optics and Photonics
- Artificial Intelligence
- Materials Science
Background:
- Diffractive neural networks (DNNs) offer energy-efficient optical AI but are limited to single-label recognition.
- Complex visual perception tasks require systems capable of multi-label recognition.
Purpose of the Study:
- To propose and validate a metasurface-based wavelength-multiplexed diffractive neural network (WMDNN).
- To enable simultaneous object category and color recognition in a single optical forward pass.
Main Methods:
- Implementation using cascaded titanium dioxide (TiO2) metasurfaces based on the Pancharatnam-Berry phase principle.
- Spatially interleaving TiO2 nanoblocks with distinct geometries for independent phase modulation across RGB wavelengths.
- Validation on a joint category-color Fashion-MNIST dataset.
Main Results:
- Achieved numerical accuracy of 91.7% and experimental accuracy of 85.8% for joint category-color recognition.
- Demonstrated simultaneous identification of object category and color in a single pass.
Conclusions:
- The proposed WMDNN architecture overcomes limitations of traditional DNNs for complex visual tasks.
- This work enables high-dimensional intelligent perception systems through wavelength multiplexing.
