A Continuous, Full-scope, Spatio-temporal Tracking Metric based on KL-divergence

TL;DR

This paper introduces a novel, continuous, parameter-free tracking metric based on KL-divergence that comprehensively evaluates object tracking performance by accounting for various error types.

Contribution

The paper proposes a unified, KL-divergence-based tracking metric recasting its components to handle different tracking errors, offering a comprehensive evaluation tool.

Findings

The metric performs well on the Oxford Town Centre Dataset.

It effectively captures false alarms, missed detections, merges, and splits.

Compared favorably to existing metrics like MOT Accuracy.

Abstract

A unified metric is given for the evaluation of object tracking systems. The metric is inspired by KL-divergence or relative entropy, which is commonly used to evaluate clustering techniques. Since tracking problems are fundamentally different from clustering, the components of KL-divergence are recast to handle various types of tracking errors (i.e., false alarms, missed detections, merges, splits). Scoring results are given on a standard tracking dataset (Oxford Town Centre Dataset), as well as several simulated scenarios. Also, this new metric is compared with several other metrics including the commonly used Multiple Object Tracking Accuracy metric. In the final section, advantages of this metric are given including the fact that it is continuous, parameter-less and comprehensive.