Temporally-Consistent Surface Reconstruction Using Metrically-Consistent Atlases
We propose a method for unsupervised reconstruction of a temporally-consistent sequence of surfaces from a sequence of time-evolving point clouds. It yields dense and semantically meaningful correspondences between frames. We represent the reconstructed surfaces as atlases computed by a neural network, which enables us to establish correspondences between frames. The key to making these correspondences semantically meaningful is to guarantee that the metric tensors computed at corresponding points are as similar as possible. We have devised an optimization strategy that makes our method robust to noise and global motions, without a priori correspondences or pre-alignment steps. As a result, our approach outperforms state-of-the-art ones on several challenging datasets.
Authors
- Jan Bednařík (ORCID: https://orcid.org/0000-0002-4018-4954)
- Noam Aigerman (ORCID: https://orcid.org/0000-0002-9116-4662)
- Shaifali Parashar (ORCID: https://orcid.org/0000-0001-9027-0503)
- Pascal Fua (ORCID: https://orcid.org/0000-0002-6702-9970)
- Vladimir G. Kim (ORCID: https://orcid.org/0000-0002-3996-6588)
- Mathieu Salzmann (ORCID: https://orcid.org/0000-0002-8347-8637)
- Siddhartha Chaudhuri
Institutions
- Google (United States) (US)
- Adobe Systems (United States) (US)
- Swiss Data Science Center (CH)
- Laboratoire d'Informatique en Images et Systèmes d'Information (FR)
- Mila - Quebec Artificial Intelligence Institute (CA)
- Institut National des Sciences Appliquées de Lyon (FR)
- École Polytechnique Fédérale de Lausanne (CH)
- Université de Montréal (CA)
Publication Details
- Journal
- IEEE Transactions on Pattern Analysis and Machine Intelligence
- Published
- 2025-02-04
- DOI
- https://doi.org/10.1109/tpami.2025.3538776
- Primary Topic
- 3D Shape Modeling and Analysis
- Type
- article
- Field-Weighted Citation Impact
- 0.00