Broadband and large-aperture metasurface edge encoders for incoherent infrared radiation.

The prevalence of computer vision systems necessitates hardware-based approaches to relieve the high computational demand of deep neural networks in resource-limited applications. One solution would be to off-load low-level image feature extraction, such as edge detection, from the digital network to the analog imaging system. To that end, this work demonstrates incoherent, broadband, low-noise optical edge detection of real-world scenes by combining the wavefront shaping of a 24-mm aperture metasurface with a refractive lens. An inverse design approach is used to optimize the metasurface for Laplacian-based edge detection across the 7.5- to 13.5-µm LWIR imaging band, allowing for facile integration with uncooled microbolometer-based LWIR imagers to encode edge information. A polarization multiplexed approach leveraging a birefringent metasurface is also demonstrated as a single-aperture implementation. This work could be applied to improve computer vision capabilities of resource-constrained systems by leveraging optical preprocessing to alleviate the computational requirements for high-accuracy image segmentation and classification.

Digital Modeling on Large Kernel Metamaterial Neural Network.

Deep neural networks (DNNs) utilized recently are physically deployed with computational units (e.g., CPUs and GPUs). Such a design might lead to a heavy computational burden, significant latency, and intensive power consumption, which are critical limitations in applications such as the Internet of Things (IoT), edge computing, and the usage of drones. Recent advances in optical computational units (e.g., metamaterial) have shed light on energy-free and light-speed neural networks. However, the digital design of the metamaterial neural network (MNN) is fundamentally limited by its physical limitations, such as precision, noise, and bandwidth during fabrication. Moreover, the unique advantages of MNN's (e.g., light-speed computation) are not fully explored via standard 3×3 convolution kernels. In this paper, we propose a novel large kernel metamaterial neural network (LMNN) that maximizes the digital capacity of the state-of-the-art (SOTA) MNN with model re-parametrization and network compression, while also considering the optical limitation explicitly. The new digital learning scheme can maximize the learning capacity of MNN while modeling the physical restrictions of meta-optic. With the proposed LMNN, the computation cost of the convolutional front-end can be offloaded into fabricated optical hardware. The experimental results on two publicly available datasets demonstrate that the optimized hybrid design improved classification accuracy while reducing computational latency. The development of the proposed LMNN is a promising step towards the ultimate goal of energy-free and light-speed AI.