S-MUSt3R: Sliding Multi-view 3D Reconstruction

Leonid Antsfeld, Boris Chidlovskii, Yohann Cabon, @vincentleroy.bsky.social Jerome Revaud
tl;dr: sliding MUSt3R for cheap long seq
The most interesting is alignment: point-based+camera-based. KDTree local desc for loop closure
arxiv.org/abs/2602.04517

S-MUSt3R: Sliding Multi-view 3D Reconstruction

Leonid Antsfeld, Boris Chidlovskii, Yohann Cabon, @vincentleroy.bsky.social, Jerome Revaud

tl;dr: sliding-window MUSt3R; overlapped input segments->MUSt3R->loop closure->PGO with iterative solver

arxiv.org/abs/2602.04517

In a new paper led by Gianluca Monaci, with @weinzaepfelp.bsky.social and myself, we explore the relationship between rel pose estimation and image goal navigation and study different architectures: late fusion, channel cat (w/ or w/o space2depth) and cross-attention.

arxiv.org/abs/2507.01667

🧵1/5

We find evidence that the agent has a plan structured on the level of paths and that its estimate of success goes beyond the effect of the next action. Abandoning a navigation option for a more promising one increases the value estimate, as the agent now expects a higher future return.
#CVPR2025