arxivcs.CVcs.AI2026-07-03
R3D: Quantitative 3D Spatial Reasoning for Egocentric Wearables
Maxwell Horton, Wei Lu, Quan Tran, Yury Astashonok, Kirmani Ahmed, Babak Damavandi, et al.
Quantitative 3D spatial reasoning from egocentric RGB-D video is a critical capability for next-generation wearable assistants. Yet existing benchmarks do not reflect the challenges of handling (1) natural egocentric video, (2) posed RGB-D video inputs, and (3) challenging quanti…