OpenVO: Open-World Visual Odometry with Temporal Dynamics Awareness
Phuc Nguyen, Anh N. NhuMing C. LinUniversity of Maryland, College Park, USA {phucnda, anhu, lin}@umd.edu https://openvo.github.io/ · 2026
Read OpenVO: Open-World Visual Odometry with Temporal Dynamics Awareness by Phuc Nguyen, Anh N. NhuMing C. LinUniversity of Maryland, College Park, USA {phucnda, anhu, lin}@umd.edu https://openvo.github.io/ (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-19035.txt.
Description
Visual odometry usually assumes fixed viewing rates and calibrated cameras, assumptions dashcam footage routinely violates. Phuc D. A. Nguyen, with two co-authors, introduces OpenVO in 2026, a frame for open-world visual tracking with time motion awareness under limited input conditions.
OpenVO estimates real-world-scale ego-motion from single-camera dashcam footage with varying viewing rates and uncalibrated cameras, allowing robust trajectory dataset building from rare driving events. Time motion data is explicitly encoded within a two-frame pose regression frame, and 3D geometric priors from foundation models are leveraged. Validation on three major autonomous-driving benchmarks shows performance gains of over twenty percent over leading approaches. Under varying viewing rates, errors fall by roughly half to over nine-tenths across metrics. OpenVO thus proves versatile for real-world 3D reconstruction and downstream uses.
Catalog record
| CREATOR | Phuc Nguyen, Anh N. NhuMing C. LinUniversity of Maryland, College Park, USA {phucnda, anhu, lin}@umd.edu https://openvo.github.io/ |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 58 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-19035