Real-Time Video Analysis of Football Matches Using YOLOv8 and Computer Vision Techniques: A Web-Based Interactive Platform

Football analytics is an integral aspect of coaching. Currently, the technology accessible in professional leagues requires pricey hardware and an establishment with multiple cameras. This paper describes the STRIKER system, an end-to-end web-based football analytics platform that is able to analyze user-submitted videos of football games and provide analytics on player tracks, speed analysis, distance analysis, team identification, movement heat map analysis, and analytical outputs using the chat interface.
STRIKER uses YOLOv8-n for the detection of players on the video, an optimized multiple-object tracking algorithm with the integration of velocity prediction and IoU association, K-mean algorithms optimized for the jersey-color-based classification of team identification, and heuristic approaches for the identification of the referee. It uses metric scaling from pixels on a standard 105-meter football ground for the estimation of the speeds of the players. Additionally, the method uses the Flask web structure with asynchronous processing.
This approach is ideal since it is able to provide analytical outputs using the chat interface with minimal web processing delay. Tests on amateur games as well as official games indicate successful detection of subjects within the video with accuracy in team identification and genuine estimations of the speeds.

Football analytics, YOLOv8, player tracking, team classification, computer vision, web-based analytics

1. Ultralytics, “YOLOv8: A New Era of Real-Time Object Detection Models [Google Scholar] [Crossref]

2. J. Bewley et al., "Simple Online and Realtime Tracking," ICIP, [Google Scholar] [Crossref]

3. N. Wojke et al., “DeepSORT: Tracking with a Deep Association Metric,” ICIP, [Google Scholar] [Crossref]

4. Z. Zhang et al., “ByteTrack,” ECC [Google Scholar] [Crossref]

5. Y. LeCun et. al. “Deep Learning.” [Google Scholar] [Crossref]

6. K. He, et al. “Deep Residual Learning for Image Recognition,” CVPR, 201 [Google Scholar] [Crossref]

7. T. Lin et al., "Microsoft COCO," ECC [Google Scholar] [Crossref]

8. A. Cunningham et al., “Automatic Team Assignment in Sports Video," IEEE TIP, 2018. [Google Scholar] [Crossref]