SeeSE3: Emergence of 3D Space in Vision Features
Quick Answer
The paper investigates whether vision foundation models represent 3D Euclidean space properties, revealing that self-supervised models exhibit strong latent correlations with 3D space.
Quick Take
The paper investigates whether vision foundation models represent 3D Euclidean space properties, revealing that self-supervised models exhibit strong latent correlations with 3D space. It introduces 'Latent-Space Navigation' techniques for visual odometry and localization, eliminating the need for explicit 3D reconstruction.
Key Points
- Probes evaluate visual feature structure against Euclidean transformations SE(3).
- Mutual neighborhood metrics measure alignment between feature neighborhoods and spatial topology.
- Self-supervised models show latent subspaces correlated with 3D Euclidean space.
- New techniques enable visual odometry in latent space without 3D reconstruction.
- Research contributes to understanding 3D awareness in vision models.
Paper Resources
📖 Reader Mode
~2 min readAbstract:In this paper, we ask whether vision foundation models construct representations that reflect the intrinsic properties of 3D Euclidean space. Unlike previous works that probe 3D awareness of vision features by regressing image-centric quantities such as depth or normals, we investigate the relation between the structure of the space of visual features and the group of Euclidean transformations $SE(3)$. We propose a set of probes to evaluate this relation from both topological and geometric perspectives: a mutual neighborhood metric that measures the alignment between feature neighborhoods and spatial topology, and a Poincaré Adapter to test the linear accessibility of the geometry of camera motion from latent displacements in static scenes. We show that self-supervised vision models, which, in principle, have not been trained with direct 3D supervision or active agency, possess latent subspaces that are remarkably strongly correlated with three-dimensional Euclidean space, when probed correctly. Building on this insight we propose a new class of "Latent-Space Navigation" techniques that perform visual odometry and localization purely in the latent space, bypassing the need for explicit 3D reconstruction.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2607.14228 [cs.CV] |
| (or arXiv:2607.14228v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2607.14228 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Maks Ovsjanikov [view email]
[v1]
Wed, 15 Jul 2026 18:00:31 UTC (8,563 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →ProMoE-FL: Prototype-conditioned Mixture of Experts for Multimodal Federated Learning with Missing Modalities
ProMoE-FL introduces a Prototype-conditioned Mixture-of-Experts framework for multimodal federated learning, effectively addressing missing modalities. It outperforms existing methods on four chest X-ray datasets, demonstrating superior feature synthesis capabilities in both homogeneous and heterogeneous settings.