Catalogue of Tools & Metrics for Trustworthy AI

These tools and metrics are designed to help AI actors develop and use trustworthy AI systems and applications that respect human rights and are fair, transparent, explainable, robust, secure and safe.

In object tracking problems (e.g., "where is the human in this image?"), Higher order tracking accuracy (HOTA) measures how well the trajectories of matching detections align, and averages this over all matching detections, while also penalising detections that don’t match. HOTA balances the effect of performing accurate detection, association and localization into a single unified metric for comparing trackers. HOTA decomposes into a family of sub-metrics which are able to evaluate each of five basic error types separately, which enables clear analysis of tracking performance.  

HOTA supports Safety by enabling the evaluation and improvement of tracking system accuracy, which is crucial for preventing harmful errors in applications like autonomous driving or security monitoring. It also supports Robustness by providing a metric to assess system performance under various conditions, helping ensure that tracking remains reliable even in challenging scenarios. However, its impact on other Trustworthy AI objectives is minimal.

Trustworthy AI Relevance

This metric addresses Robustness and Transparency by quantifying relevant system properties. Robustness: HOTA explicitly measures association accuracy across frames in addition to detection performance, so it evaluates how consistently a tracking system maintains correct identities under adverse conditions (occlusions, crowded scenes, noise, motion), making it a direct proxy for the system's resilience and reliability. Transparency: HOTA is decomposable into detection and association components (e.g., DetA and AssA), providing interpretable diagnostics that reveal whether failures stem from detection or from identity association.

About the metric




Lifecycle stage(s):


Target users:


Risk management stage(s):


Github stars:

  • 12172

Github forks:

  • 1849

Modify this metric

Partnership on AI

Disclaimer: The tools and metrics featured herein are solely those of the originating authors and are not vetted or endorsed by the OECD or its member countries. The Organisation cannot be held responsible for possible issues resulting from the posting of links to third parties' tools and metrics on this catalogue. More on the methodology can be found at https://oecd.ai/catalogue/faq.