Related Experiment Video
Updated: Dec 26, 2025
14:02
Measuring Engagement of Spectators of Social Digital Games
Published on: July 3, 2021
3.8K
Active Player Detection in Handball Scenes Based on Activity Measures
Miran Pobar1, Marina Ivasic-Kos1
1Department of Informatics University of Rijeka, Rijeka 51000, Croatia.
Sensors (Basel, Switzerland)
|March 19, 2020
Summary
This study introduces a novel method for detecting active players in team sports using object detection and activity tracking. The system accurately identifies key players performing specific handball techniques in complex, cluttered environments.
Area of Science:
- Computer Vision
- Sports Analytics
- Machine Learning
Background:
- Team sports training involves complex scenes with multiple players and actions.
- Identifying the most active player performing specific techniques is challenging due to scene clutter.
Purpose of the Study:
- To develop a robust method for detecting and tracking active players in handball.
- To determine the most active player executing a given handball technique.
Main Methods:
- Combined YOLO object detection with activity measures (optical flow, STIPs, CNNs) and tracking algorithms (Hungarian, Deep SORT).
- Proposed novel activity measures and an evaluation metric for active player detection.
Main Results:
- Successfully tested the method on custom handball and existing basketball datasets.
- Demonstrated effective detection and tracking of active players in challenging, real-world scenarios.
Conclusions:
- The proposed active player detection method effectively handles cluttered team sports scenes.
- The integration of object detection, activity analysis, and tracking provides accurate player identification.
