Event Stream-based Visual Object Tracking: HDETrack V2 and A High-Definition Benchmark

TL;DR

This paper introduces HDETrack V2, a high-definition event-based tracking method utilizing hierarchical knowledge distillation, temporal Fourier transforms, and test-time tuning, validated on a new high-resolution dataset, EventVOT.

Contribution

The paper presents a novel high-resolution event-based tracking dataset and an advanced Transformer-based tracking method with hierarchical knowledge distillation and test-time tuning.

Findings

Effective tracking on high-resolution event data

Superior performance on multiple datasets

Validated robustness and accuracy

Abstract

We then introduce a novel hierarchical knowledge distillation strategy that incorporates the similarity matrix, feature representation, and response map-based distillation to guide the learning of the student Transformer network. We also enhance the model's ability to capture temporal dependencies by applying the temporal Fourier transform to establish temporal relationships between video frames. We adapt the network model to specific target objects during testing via a newly proposed test-time tuning strategy to achieve high performance and flexibility in target tracking. Recognizing the limitations of existing event-based tracking datasets, which are predominantly low-resolution, we propose EventVOT, the first large-scale high-resolution event-based tracking dataset. It comprises 1141 videos spanning diverse categories such as pedestrians, vehicles, UAVs, ping pong, etc. Extensive experiments on both low-resolution (FE240hz, VisEvent, FELT), and our newly proposed high-resolution EventVOT dataset fully validated the effectiveness of our proposed method. Both the benchmark dataset and source code have been released on https://github.com/Event-AHU/EventVOT_Benchmark