ReViV proposes a unified framework that reconstructs both viewer motion and surrounding 3D scene from a single egocentric RGB video, without needing pre-computed camera trajectories. By jointly modeling scene perception…

#egocentricAI #computervision #wearableAI
https://arxiv.org/abs/2607.17790