ViT-BF: vision transformer with border-aware features for visual tracking

doi:10.1007/s00371-025-03964-z

ViT-BF: vision transformer with border-aware features for visual tracking

2025-05-25

0

PRE

AI

DOI：10.1007/s00371-025-03964-z

Original

Original request for help

Save

Abstract

En 中文

Existing object trackers commonly approach the tracking process by utilizing classification and regression techniques. However, they often encounter difficulties in managing complex scenarios such as occlusions and appearance changes. Moreover, the quality of candidate boxes is a critical factor for affecting the tracking performance. To overcome these challenges, this study introduces a border-aware tracking framework based on a vision transformer (ViT), termed ViT-BF. Through the integration of a boundary alignment operation, ViT-BF extracts boundary features from the extremal points of objects, thereby enhancing classification and regression precision. To handle the dynamic appearance variations of objects, ViT-BF integrates a template update mechanism through a score prediction module (SPM), which enhances the tracker's robustness and accuracy. Experimental results reveal that ViT-BF demonstrates excellent performance across multiple benchmarks, including an AUC score of 85.0% on TrackingNet, 89.4% in normalized precision, and 84.4% in precision. The promising results extend to other standard datasets such as LaSOT, GOT-10k, and UAV123, validating our method's strong stability and adaptability in diverse tracking scenarios.

Keywords:

Visual tracking

border features

anchor-free tracker

boundary alignment

ViT-BF: vision transformer with border-aware features for visual tracking

Abstract

Journal

Organization