Action Spotting

SoccerNet Action Spotting

In order to understand the salient actions of a broadcast soccer game, SoccerNet introduces the task of action spotting, which consists in finding all the actions occurring in the videos. Beyond soccer understanding, this task addresses the more general problem of retrieving moments with a specific semantic meaning in long untrimmed videos.

Our task.

Action spotting consists in localizing when and which soccer action occurs, in our dataset among 17 classes. Each action is annotated with a single timestamp, making those annotations quite scattered in our long videos. 

All of our classes.

We provide annotation for all common soccer actions: {Penalty, Kick-off, Goal, Substitution, Offside, Shots on target,  Shots off target, Clearance, Ball out of play, Throw-in, Foul, Indirect free-kick, Direct free-kick, Corner, Yellow card, Red card, Yellow->red card}.

Our data.

The data consists of 500 videos from soccer broadcast games available at two resolutions (720p and 224p). We also provide extracted features at 2 frames per second for an easier use, including the feature used by the 2021 challenge winners, Baidu Research.  The challenge set is composed of 50 separate games.

Our Metric.

Traditionally, the average-mAP as defined in this original paper was used as the main metric. However, starting in 2022, we introduced the tight average-mAP, which is the same metric, but with a smaller temporal tolerance ([1-5] seconds instead of [5-60 seconds]).

For more details, check out our development kit on github

2023 Challenge leaderboard

Published research leaderboard

2022 challenge leaderboard

2021 challenge leaderboard

If your work is missing from one of these tables, don't hesitate to contact us through one of our social media.

Our videos on Action Spotting

SoccerNet-v2: a new MASSIVE soccer dataset

In this video, we present our paper: “SoccerNet-v2: A Dataset and Benchmarks for Holistic Understanding of Broadcast Soccer Videos” published at the CVPR 2021 workshop CVsports. We provide 300,000 temporal annotations within 500 soccer games. This allows a 17-class action spotting task, a 13-class camera boundary detection task, and a novel replay grounding task. We provide benchmarks for all these tasks to start an international challenge.

A Context-Aware Loss Function for Action Spotting in Soccer Videos

In this video, we present our paper: “A Context-Aware Loss Function for Action Spotting in Soccer Videos” published at CVPR 2020. This work introduces a novel loss to gather the temporal context surrounding the actions, further used to spot those actions in soccer videos. We achieve state-of-the-art performances on SoccerNet and improvements on ActivityNet!

SoccerNet-v2: Our new soccer dataset and AI challenge

SoccerNet-v2 challenge - Tutorial #2 (live session)

SoccerNet-v2 challenge - Tutorial #3 (live session)


SoccerNet-v2 challenge - Tutorial #4 part 1 (live session) ft. Matteo Tomei

SoccerNet-v2 challenge - Tutorial #4 part 2 (live session) ft. Bastien Vanderplaetse

SoccerNet-v2 challenge - Tutorial #5 part 1 (live session) ft. Kanav Vats

How to cite this work ?

@InProceedings{Deliège2020SoccerNetv2,

      title={SoccerNet-v2 : A Dataset and Benchmarks for Holistic Understanding of Broadcast Soccer Videos}, 

      author={Adrien Deliège and Anthony Cioppa and Silvio Giancola and Meisam J. Seikavandi and Jacob V. Dueholm and Kamal Nasrollahi and Bernard Ghanem and Thomas B. Moeslund and Marc Van Droogenbroeck},

      year={2021},

      booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},

      month = {June},