GTATrack: Winner Solution to SoccerTrack 2025 with Deep-EIoU and Global Tracklet Association Conference: Proceedings of the 8th International ACM Workshop on Multimedia Content Analysis in Sports; October 27–28, 2025; Dublin, Ireland Proceedings of the 8th International ACM Workshop on Multimedia Content Analysis in Sports (MMSports ’25), October 27–28, 2025, Dublin, Ireland DOI: 10.1145/3728423.3759416 ISBN: 979-8-4007-1835-9/2025/10 mspt43 CCS: Computing methodologies Computer vision tasks CCS: Computing methodologies Active vision CCS: Computing methodologies Activity recognition and understanding

Rong-Lin Jian, Ming-Chi Luo, Chen-Wei Huang, Chia-Ming Lee, Yu-Fan Lin, Chih-Chung Hsu · 2026

Read GTATrack: Winner Solution to SoccerTrack 2025 with Deep-EIoU and Global Tracklet Association Conference: Proceedings of the 8th International ACM Workshop on Multimedia Content Analysis in Sports; October 27–28, 2025; Dublin, Ireland Proceedings of the 8th International ACM Workshop on Multimedia Content Analysis in Sports (MMSports ’25), October 27–28, 2025, Dublin, Ireland DOI: 10.1145/3728423.3759416 ISBN: 979-8-4007-1835-9/2025/10 mspt43 CCS: Computing methodologies Computer vision tasks CCS: Computing methodologies Active vision CCS: Computing methodologies Activity recognition and understanding by Rong-Lin Jian, Ming-Chi Luo, Chen-Wei Huang, Chia-Ming Lee, Yu-Fan Lin, Chih-Chung Hsu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-00484.txt.

Description

Rong-Lin Jian and colleagues present GTATrack, the winning entry in the SoccerTrack Challenge 2025 for tracking players in fisheye soccer video. Irregular motion, matching kits, frequent occlusions, and fisheye distortion with extreme scale change make the task hard. The hierarchical framework pairs Deep-EIoU online association, which makes no motion assumptions, with global tracklet association for trajectory-level refinement. Pseudo-labelling lifts detector recall on small and distorted targets, cutting identity switches and fragmentation, and the method reached a HOTA score of 0.60 with false positives down to 982.

Catalog record

CREATORRong-Lin Jian, Ming-Chi Luo, Chen-Wei Huang, Chia-Ming Lee, Yu-Fan Lin, Chih-Chung Hsu
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31186
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602.00484

Browse the full library · About the Tower of Babel