Work

Learned 3D depth estimation and object detection via an integrated metasystem with wide field of view

Year: 2026
Type: article
Abstract: Object detection and 3D depth estimation have been widely adopted in autonomous driving, intelligent surveillance, augmented reality, and industrial automation, providing core support for scene perception, object-level interaction, and spatial understanding. Metasurfaces have recently risen to prominence in optical research, providing unique functionalities for imaging and 3D vision, while keeping device dimensions small. However, these meta-based 3D imaging methods mainly rely on binocular or multi-view vision, leading to an incompressible baseline dimension, especially working in the long distance. Here, we introduce an optimized metalens design and monocular depth estimation network for object detection and 3D depth recovery. By using the inverse-design metalens of wide field of view integrated with a commercial near-infrared sensor, we experimentally demonstrate the snapshot acquisition, high depth resolution, and accurate object detection ability. Additionally, we showcase the robustness of our approach by applying it to people and car detection in autonomous driving, highlighting its potential for deployment in embodied AI and industrial monitoring.
Language: English
Open Access status: gold
Estimated APC paid: $2,270
OpenAlex ID: W7197042910
FWCI:
Cites: 27
Cited by:
Related to: 10