Researchers at  the MIT Media Lab have demonstrated that consumer-grade LiDAR sensors—already embedded in many smartphones—could enable non-line-of-sight (NLOS) imaging: the ability to detect and track objects outside a camera's direct field of view.

Featured on the Nature Podcast, the work moves beyond earlier lab demonstrations that relied on high-powered lasers, instead leveraging low-power LiDAR found in everyday devices. The key insight is motion. By simultaneously modeling object shape, object position, and camera position over time—what the team calls an aperture sampling model—the system enforces temporal coherence across sequential measurements, extracting meaningful signal from noisy data.

The team demonstrated four capabilities: object tracking, shape reconstruction, multi-object tracking, and camera self-localization using hidden landmarks—particularly promising for robotics.

Outputs are not photographs, but progressively richer inferences about presence, motion, and shape. Researcher Siddharth Somasundaram and colleagues have made their code publicly available, with sensors assemblable for under $50.