CoWTracker: Tracking by Warping instead of Correlation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Lai, Zihang, Insafutdinov, Eldar, Sucar, Edgar, Vedaldi, Andrea
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866917248319356928
author Lai, Zihang
Insafutdinov, Eldar
Sucar, Edgar
Vedaldi, Andrea
author_facet Lai, Zihang
Insafutdinov, Eldar
Sucar, Edgar
Vedaldi, Andrea
contents Dense point tracking is a fundamental problem in computer vision, with applications ranging from video analysis to robotic manipulation. State-of-the-art trackers typically rely on cost volumes to match features across frames, but this approach incurs quadratic complexity in spatial resolution, limiting scalability and efficiency. In this paper, we propose \method, a novel dense point tracker that eschews cost volumes in favor of warping. Inspired by recent advances in optical flow, our approach iteratively refines track estimates by warping features from the target frame to the query frame based on the current estimate. Combined with a transformer architecture that performs joint spatiotemporal reasoning across all tracks, our design establishes long-range correspondences without computing feature correlations. Our model is simple and achieves state-of-the-art performance on standard dense point tracking benchmarks, including TAP-Vid-DAVIS, TAP-Vid-Kinetics, and Robo-TAP. Remarkably, the model also excels at optical flow, sometimes outperforming specialized methods on the Sintel, KITTI, and Spring benchmarks. These results suggest that warping-based architectures can unify dense point tracking and optical flow estimation.
format Preprint
id arxiv_https___arxiv_org_abs_2602_04877
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle CoWTracker: Tracking by Warping instead of Correlation
Lai, Zihang
Insafutdinov, Eldar
Sucar, Edgar
Vedaldi, Andrea
Computer Vision and Pattern Recognition
Dense point tracking is a fundamental problem in computer vision, with applications ranging from video analysis to robotic manipulation. State-of-the-art trackers typically rely on cost volumes to match features across frames, but this approach incurs quadratic complexity in spatial resolution, limiting scalability and efficiency. In this paper, we propose \method, a novel dense point tracker that eschews cost volumes in favor of warping. Inspired by recent advances in optical flow, our approach iteratively refines track estimates by warping features from the target frame to the query frame based on the current estimate. Combined with a transformer architecture that performs joint spatiotemporal reasoning across all tracks, our design establishes long-range correspondences without computing feature correlations. Our model is simple and achieves state-of-the-art performance on standard dense point tracking benchmarks, including TAP-Vid-DAVIS, TAP-Vid-Kinetics, and Robo-TAP. Remarkably, the model also excels at optical flow, sometimes outperforming specialized methods on the Sintel, KITTI, and Spring benchmarks. These results suggest that warping-based architectures can unify dense point tracking and optical flow estimation.
title CoWTracker: Tracking by Warping instead of Correlation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2602.04877