SoccerNet
Soccer Video Understanding Benchmark Suite
What is SoccerNet?
SoccerNet is a large-scale dataset for soccer video understanding. It has evolved over the years to include various tasks such as action spotting, camera calibration, player re-identification and tracking. It is composed of 550 complete broadcast soccer games and 12 single camera games taken from the major European leagues. SoccerNet is not only dataset, but also yearly challenges where the best teams compete at the international level. Join the SoccerNet community on Discord to discover how SoccerNet can further push your own research and industry.
News
[2023-02] Announcing the SoccerNet Challenge 2023, featured at CVPR'23! We have introduced 7 challenges including Dense Video Captioning, Ball Action Spotting, Action Spotting, Camera Calibration, Tracking, ReID and Jersey Number Recognition! Join our Discord Channel for more information.
[2022-10] The SoccerNet team collaborated with AinSports and won the 2nd place of the Thakaa competition in AI+Sports!
[2022-10] The results of the SoccerNet 2022 Challenge were presented at the ACM Workshop on Multimedia Content Analysis in Sports.
[2022-10] The latest SoccerNet activities in Action Spotting were highlighted during the FIFA Research Symposium 2022.
[2022-08] Our SoccerNet 2022 Challenge results paper is available here.
[2022-08] The leaderboards for the SoccerNet 2022 Challenge are available on the task pages!
[2022-06] Our SoccerNet-Tracking paper won the Best Paper Award at the CVSports workshop!
[2022-06] SoccerNet challenge results were announced during the ActivityNet and CVSports workshop at CVPR. Leaderboards and recordings coming soon!
[2022-05] Replay of the first tutorial available on YouTube: https://youtu.be/JtoziK5H-pw
[2022-04] Live tutorials on Zoom (https://kaust.zoom.us/j/99247680793) the 27 and 28 of April 2022 from 4pm to 7pm CEST.
[2022-03] The tracking data and challenge is now available. Check out the task and how to download the data.
[2022-01] Announcing the SoccerNet Challenge 2022, featured at CVPR'22! We have introduced 4 challenges including Calibration, Tracking, ReID and Spotting! Join our Discord Channel for more information.
[2021-06] The SoccerNet-v2 paper won the Best Paper Award at CVSports during CVPR2021! More info on Twitter.
[2021-06] The winner of the SoccerNet-v2 Challenge sponsored by Second Spectrum for both Action Spotting and Replay Grounding is Baidu Research Vidpress Sports.
[2021-06] The SoccerNet-v2 Challenge will be held on Thursday June 17th, 2pm UTC on Zoom. Recording available on Youtube. Join us on Discord for further information.
[2021-04] SoccerNet-v2 accepted in CVSports'21, aside with 2 follow up papers: Temporally-Aware Feature Pooling (NetVLAD++) and Camera Calibration with Player Detection.
[2021-03] The SoccerNet Challenge is up and running! We will be hosting the SoccerNet Challenge in the CVPR'21 International Challenge on Activity Recognition Workshop. Participate and win up to 1000 USD from our sponsor Second Spectrum!
[2020-06] Presentation of Context-Aware Loss Function (CALF) for Action Spotting during CVPR'20.
[2018-06] Official release of SoccerNet with 500 games for Action Spotting during CVSports'18.
SoccerNet at a glance
550 Broadcast Games
12 Single Camera Games
> 5M Feature Frames
> 300k Temporal Annotations
> 1M Spatial Annotations
6 Challenging Task
Latest YouTube Videos
Soccer Player Tracking, Re-ID, Camera Calibration and Action Spotting - SoccerNet Challenges 2022
In this video, we present our new SoccerNet Challenges for CVPR 2022! We introduce the three tasks of Calibration, Re-identification and Tracking on soccer games, in partnership with EVS Broadcast Equipment, SportRadar and Baidu Research. We also reiterate our previous Action Spotting and Replay Grounding Challenges at the ActivityNet workshop. Good luck fighting for the throne!
Camera Calibration and Player Localization in SoccerNet-v2
In this video, we present our paper: “Camera Calibration and Player Localization in SoccerNet-v2 and Investigation of their Representations for Action Spotting” published at the CVPR 2021 workshop CVsports. This work introduces a camera calibration and player localization algorithm for the SoccerNet-v2 dataset. Next, a method based on CALF is proposed to investigate different types of representations of this player localization for the task of action spotting.