DAL -- A Deep Depth-aware Long-term Tracker

TL;DR

This paper introduces a deep depth-aware long-term tracker that combines high accuracy and speed for RGBD tracking by embedding depth information into deep features and utilizing a depth-aware correlation filter.

Contribution

It reformulates deep discriminative correlation filters to incorporate depth data, enabling fast and accurate long-term RGBD tracking.

Findings

Achieves state-of-the-art performance on multiple RGBD benchmarks.

Runs at 20 frames per second, balancing speed and accuracy.

Outperforms existing RGB and RGBD trackers in accuracy.

Abstract

The best RGBD trackers provide high accuracy but are slow to run. On the other hand, the best RGB trackers are fast but clearly inferior on the RGBD datasets. In this work, we propose a deep depth-aware long-term tracker that achieves state-of-the-art RGBD tracking performance and is fast to run. We reformulate deep discriminative correlation filter (DCF) to embed the depth information into deep features. Moreover, the same depth-aware correlation filter is used for target re-detection. Comprehensive evaluations show that the proposed tracker achieves state-of-the-art performance on the Princeton RGBD, STC, and the newly-released CDTB benchmarks and runs 20 fps.