FlowEgoMotion: Scene-Conditioned Rectified Flow for Egocentric Full-Body Motion Estimation

Estimating full-body motion from a head-mounted egocentric device is severely underconstrained: visual–inertial SLAM gives reliable egomotion estimation, but most of the wearer's body is never directly visible, and prior diffusion-based approaches incur high inference cost while underusing scene context. We present FlowEgoMotion, a scene-conditioned rectified-flow model that combines the SLAM head trajectory with two complementary scene pathways – time-aligned local geometry fused into the egomotion stream, and global tokens encoding room-scale structure – to condition a Transformer that predicts a motion vector field, while global root motion is recovered analytically from the trajectory. On Nymeria, a controlled comparison with the same backbone shows that scene conditioning reduces MPJPE from $86.7$ to $75.6$ mm, Wrist-PE from $224.8$ to $147.3$ mm, and floor penetration from $13.5$ to $0.5$ mm, and the gains carry over to ADT without fine-tuning. FlowEgoMotion also improves on the closest published scene-aware system, HMD${}^2$, on the same benchmark, reducing MPJPE from $83.6$ to $75.6$ mm, Wrist-PE from $166.4$ to $147.3$ mm, and floor penetration from $10.3$ to $0.5$ mm without using the RGB features that HMD${}^2$ relies on. With only two flow steps, it needs about $12.5$ ms of GPU compute per second of generated motion and is more accurate than a matched 30-step diffusion model at $15\times$ lower cost.