A team of researchers from Peking University, Beijing Institute of Technology, and the Chinese Academy of Sciences has developed a new optical device capable of asymmetrically transmitting and transforming images in two opposite directions. Dubbed the “Janus meta-imager,” the device is based on diffractive deep neural networks (D²NNs) and metasurfaces, and operates without external power or nonlinear materials.
Unlike conventional asymmetric imaging systems that rely on active components or limited unidirectional functions, the Janus meta-imager can perform different imaging tasks in forward and backward directions. For example, in simulations, input letters are directly transmitted forward, while simultaneously being transformed into clothing icons backward.
The device consists of cascaded metasurface layers optimized via deep learning algorithms. Experiments in the near-infrared band (800 nm) validated its asymmetric imaging function with high correlation to target outputs. The imager operates at about 10 kHz frame rates, consumes minimal energy, and measures only 0.5 × 0.5 mm².
Potential applications include all-optical encryption, where the device can switch between plaintext transmission and ciphertext generation depending on the direction of light, and high-density optical data storage with direction-multiplexed information.
PhotoniX
Experimental study
Not applicable
Janus meta-imager: asymmetric image transmission and transformation enabled by diffractive neural networks
24-Dec-2025
The authors declare no competing interests.