Building the largest and most comprehensive open-source dataset and benchmark for sports video analysis

Artificial intelligence (AI) holds immense potential to transform our understanding and experience of sports. At the forefront of this transformation is SoccerNet, the largest and most comprehensive open-source dataset and benchmark for sports video analysis. Launched in 2018, SoccerNet is a collaboration led by KAUST (Saudi Arabia), and ULiège (Belgium). Nowadays, SoccerNet has grown into a truly international project with contributions from academic institutions such as Aalborg University (Denmark), UCLouvain (Belgium), and EPFL (Switzerland), as well as industry leaders like EVS Broadcast Equipment (Belgium), Sportradar (Switzerland), Baidu Research (USA), and NASK Science (Poland).
SoccerNet addresses a critical challenge in sports AI research: the lack of large, publicly available datasets that enable meaningful comparison across different AI approaches. By providing more than 850 hours of videos, all manually annotated with event and player data, SoccerNet offers a unique resource for researchers worldwide. The dataset is provided in open access to researchers, and offers a complete framework to explore methods for tasks such as action spotting, camera calibration, or player tracking, fostering the development of state-of-the-art solutions in computer vision.
Beyond the dataset, SoccerNet has catalyzed a vibrant global research community, with over a thousand researchers engaging through its Discord platform. The annual SoccerNet challenges have attracted 1,500 submissions from 134 teams, including prestigious institutions like Stanford and Amazon, further advancing the field of sports video understanding. This collaborative effort has resulted in over 20 published papers at ULiège and opened doors to partnerships with key organizations such as FIFA.
The success of SoccerNet underscores the importance of providing open data and benchmarks to drive research innovation. The project's impact demonstrates how strategic investments in open science can yield significant returns in research output, international recognition, and collaborative opportunities for participating institutions.