Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Deep HM-SORT: Enhancing Multi-Object Tracking in Sports with Deep Features, Harmonic Mean, and Expansion IOU

Matias Gran-Henriksen, Hans Andreas Lindgaard, Gabriel Kiss, Frank Lindseth

TL;DR

Deep HM-SORT tackles online multi-object tracking in sports by fusing motion and appearance through a harmonic-mean cost, improving robustness when players look similar and move erratically. It also preserves all tracklets indefinitely to strengthen re-identification after players exit and re-enter the frame. Empirical results show state-of-the-art HOTA on SportsMOT (80.1) and competitive performance on SoccerNet Tracking Challenge 2023 (85.4), with substantial reductions in ID swaps and strong overall metrics, while maintaining online processing. The approach advances automated sports analytics by delivering higher accuracy and reliability without adding computational cost, and it can be integrated with other tracking-by-detection methods to enhance ID consistency in sports broadcasting and analytics.

Abstract

This paper introduces Deep HM-SORT, a novel online multi-object tracking algorithm specifically designed to enhance the tracking of athletes in sports scenarios. Traditional multi-object tracking methods often struggle with sports environments due to the similar appearances of players, irregular and unpredictable movements, and significant camera motion. Deep HM-SORT addresses these challenges by integrating deep features, harmonic mean, and Expansion IOU. By leveraging the harmonic mean, our method effectively balances appearance and motion cues, significantly reducing ID-swaps. Additionally, our approach retains all tracklets indefinitely, improving the re-identification of players who leave and re-enter the frame. Experimental results demonstrate that Deep HM-SORT achieves state-of-the-art performance on two large-scale public benchmarks, SportsMOT and SoccerNet Tracking Challenge 2023. Specifically, our method achieves 80.1 HOTA on the SportsMOT dataset and 85.4 HOTA on the SoccerNet-Tracking dataset, outperforming existing trackers in key metrics such as HOTA, IDF1, AssA, and MOTA. This robust solution provides enhanced accuracy and reliability for automated sports analytics, offering significant improvements over previous methods without introducing additional computational cost.

Deep HM-SORT: Enhancing Multi-Object Tracking in Sports with Deep Features, Harmonic Mean, and Expansion IOU

TL;DR

Deep HM-SORT tackles online multi-object tracking in sports by fusing motion and appearance through a harmonic-mean cost, improving robustness when players look similar and move erratically. It also preserves all tracklets indefinitely to strengthen re-identification after players exit and re-enter the frame. Empirical results show state-of-the-art HOTA on SportsMOT (80.1) and competitive performance on SoccerNet Tracking Challenge 2023 (85.4), with substantial reductions in ID swaps and strong overall metrics, while maintaining online processing. The approach advances automated sports analytics by delivering higher accuracy and reliability without adding computational cost, and it can be integrated with other tracking-by-detection methods to enhance ID consistency in sports broadcasting and analytics.

Abstract

This paper introduces Deep HM-SORT, a novel online multi-object tracking algorithm specifically designed to enhance the tracking of athletes in sports scenarios. Traditional multi-object tracking methods often struggle with sports environments due to the similar appearances of players, irregular and unpredictable movements, and significant camera motion. Deep HM-SORT addresses these challenges by integrating deep features, harmonic mean, and Expansion IOU. By leveraging the harmonic mean, our method effectively balances appearance and motion cues, significantly reducing ID-swaps. Additionally, our approach retains all tracklets indefinitely, improving the re-identification of players who leave and re-enter the frame. Experimental results demonstrate that Deep HM-SORT achieves state-of-the-art performance on two large-scale public benchmarks, SportsMOT and SoccerNet Tracking Challenge 2023. Specifically, our method achieves 80.1 HOTA on the SportsMOT dataset and 85.4 HOTA on the SoccerNet-Tracking dataset, outperforming existing trackers in key metrics such as HOTA, IDF1, AssA, and MOTA. This robust solution provides enhanced accuracy and reliability for automated sports analytics, offering significant improvements over previous methods without introducing additional computational cost.
Paper Structure (16 sections, 4 equations, 2 figures, 4 tables)

This paper contains 16 sections, 4 equations, 2 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Multiple Object Tracking using Kalman Filter
  4. Intersection Over Union
  5. Appearance-based Multiple Object Tracking
  6. Multi-Object Tracking in Sports
  7. Proposed Method
  8. IoU - Re-ID Fusion
  9. Tracklet Management
  10. Experiments
  11. Experimental Settings
  12. Ablation Study
  13. Benchmarks Evaluation
  14. Limitations
  15. Conclusion
  16. ...and 1 more sections

Figures (2)

  • Figure 1: Showing the IoU between two expanded bounding boxes. Source CBIOU.
  • Figure 2: Demonstrating two similar players from the SoccerNet dataset soccerNetPaper before they are about to cross each other, potentially leading to an ID-swap.