AI & ML New Capability

A training-free operator for streaming 3D reconstruction reduces geometric drift using Grassmannian manifolds.

arXiv · March 17, 2026 · 2603.14765

Hui Deng, Yuxin Mao, Yuxin He, Yuchao Dai

The Takeaway

SSR offers a plug-and-play solution to the drift problem in long-horizon 3D reconstruction without requiring model retraining. It is particularly valuable for real-time applications in robotics and augmented reality where latency and temporal coherence are critical.

From the abstract

Streaming 3D reconstruction demands long-horizon state updates under strict latency constraints, yet stateful recurrent models often suffer from geometric drift as errors accumulate over time. We revisit this problem from a Grassmannian manifold perspective: the latent persistent state can be viewed as a subspace representation, i.e., a point evolving on a Grassmannian manifold, where temporal coherence implies the state trajectory should remain on (or near) thisthis http URLon this view, we pro