aiqbench analysis
This research introduces Poincar3, a self-supervised framework that decouples geometric reasoning from visual appearance by utilizing self-distillation across multiple views rather than relying on RGB reconstruction. By training without explicit 3D supervision, the method achieves superior performance in correspondence estimation and camera pose estimation compared to leading models like DINOv3. The study demonstrates that learned features can accurately encode camera motion through a lightweight adapter, advancing the state of multi-view visual representation learning.
Emergent Multi-View Geometry Through Self-Distillation
arXiv:2609.39227v1 Announce Type: cross Abstract: Over a century ago, Henri Poincar\'e argued that a motionless observer cannot acquire the notion of space. Yet, most visual representation learning methods operate on individual images, while those that leverage multiple views rely on RGB reconstruction, entangling geometry with appearance. We propose Poincar3, a self-supervised method that learns representations from multiple views through self-distillation instead of RGB reconstruction. We combine masked patch and image-level distillation with a teacher that observes additional views, enabling training from scratch without explicit 3D supervision. Poincar3 outperforms both previous single and multi-view self-supervised approaches such as DINOv3, MuM, and Muskie on correspondence estimation, camera pose estimation, and 3D reconstruction. Using a lightweight Poincar\'e adapter, we also find that our learned features encode camera motion more accurately than existing self-supervised representations.
Abstract:Over a century ago, Henri Poincaré argued that a motionless observer cannot acquire the notion of space. Yet, most visual representation learning methods operate on individual images, while those that leverage multiple views rely on RGB reconstruction, entangling geometry with appearance. We propose Poincar3, a self-supervised method that learns representations from multiple views through self-distillation instead of RGB reconstruction. We combine masked patch and image-level distillation with a teacher that observes additional views, enabling training from scratch without explicit 3D supervision. Poincar3 outperforms both previous single and multi-view self-supervised approaches such as DINOv3, MuM, and Muskie on correspondence estimation, camera pose estimation, and 3D reconstruction. Using a lightweight Poincaré adapter, we also find that our learned features encode camera motion more accurately than existing self-supervised representations.
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
arXiv:2609.39227 [cs.CV]
(or arXiv:2609.39227v1 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2609.39227 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 226 words
Related coverage
[AINews] Reality Checks on AI News (Yegge shuts down Gas Town, Databricks’ +60% Astra cost)
Latent Space · Sep 17, 2026
Planning with the Views via Scene Self-Exploration
arXiv · Jun 16, 2026
[AINews] not much happened today
Latent Space · Sep 10, 2026
Planning with the Views
arXiv · Jun 17, 2026