Neural nanophotonic object detector with an ultrawide field-of-view

Intelligent object detection, which extracts crucial information such as object categories and locations, plays a vital role in emerging technologies, including autonomous driving, the Internet of Things, and mobile sensing. Achieving high detection performance over a wide angular range typically demands complex and bulky optical systems, significantly limiting the detectors’ applicability. Here, we present an ultracompact neural nanophotonic object detector based on a metalens array, achieving high-accuracy object recognition (>96%) and high-precision localization (<0.1 deg) across an ultrawide field-of-view (>135 deg) under static conditions. Benefiting from its high compactness and planar architecture, we integrated the device into a miniature unmanned aerial vehicle, enabling real-time intelligent detection across diverse real-world environments with exceptional flexibility and mobility. Our work fully leverages the ultrathin and lightweight advantages of the metalens and demonstrates a full-stack design-to-system pipeline, spanning nanostructure fabrication, neural algorithm development, and system-level integration, offering a scalable framework for next-generation intelligent photonic sensing systems.