OpenSportsLab/soccernetpro-classification-GAR

SoccerNet-GAR: Pixels or Positions? Benchmarking Modalities in Group Activity Recognition

SoccerNet-GAR is a large-scale multimodal dataset for Group Activity Recognition (GAR) built from all 64 matches of the FIFA World Cup 2022 tournament. It provides synchronized broadcast video and player tracking data for 94,285 annotated group activities across 10 action classes, enabling direct comparison between video-based and tracking-based approaches.

Dataset Details

Description

SoccerNet-GAR is the first dataset to provide synchronized tracking and video modalities for the same action instances in group activity recognition. For each annotated event, a 4.5-second temporal window is extracted from both the broadcast video and player tracking streams, centered on the event timestamp. Within each window, 16 samples are taken at 30 fps with a 9-frame interval.

The dataset contains two input modalities:

• Video Modality: Broadcast footage at 720p resolution, including multi-view broadcast cameras. Each frame is part of a temporal sequence sampled within the event window, capturing appearance cues, scene context, and visual motion patterns.
• Tracking Modality: 2D player positions and 3D ball coordinates sampled at 30 fps, automatically extracted from broadcast footage and manually refined by annotators. Player positions span x in [-60, 60]m, y in [-42, 41]m; ball positions include height z in [-8, 25]m. Each entity state encodes spatial coordinates, entity identity (one-hot encoding), and motion dynamics (displacement vectors between consecutive frames).

Property	Value
Curated by	SoccerNet Team (KAUST, University of Liege)
Original Data Source	Gradient Sports (formerly PFF FC)
Total Events	94,285
Matches	64 (FIFA World Cup 2022)
Action Classes	10
Modalities	Video + Tracking
Avg. Events per Match	1,473

Sources

• Repository: https://github.com/drishyakarki/pixels_vs_positions
• Paper: Pixels or Positions? Benchmarking Modalities in Group Activity Recognition (arXiv:2511.12606)

Dataset Structure

Action Classes

The dataset contains 10 action classes reflecting common football events:

Class	Count	Proportion
PASS	59,657	63.3%
TACKLE	11,107	11.8%
OUT	6,389	6.8%
HEADER	5,803	6.2%
HIGH PASS	2,697	2.9%
THROW IN	2,618	2.8%
CROSS	2,412	2.6%
FREE KICK	1,827	1.9%
SHOT	1,559	1.7%
GOAL	216	0.2%

The dataset exhibits severe class imbalance (276:1 ratio between PASS and GOAL), reflecting the natural distribution of football events.

Splits

Data is split at the match level to prevent leakage:

Split	Matches	Events	Proportion
Train	45	66,901	71.0%
Validation	9	12,865	13.6%
Test	10	14,519	15.4%

Branches

This repository is organized into the following branches:

Branch	Contents
main	Dataset card and documentation.
paper-data	The exact dataset needed to reproduce the results in the paper. Contains broadcast videos (1 npy clip per event) and tracking files (1 parquet file per full match).
frames	1 npy clip per event for the video modality. Annotations are in SoccerNetPro format.
tracking-parquet	1 parquet clip per event for the tracking modality. Annotations are in SoccerNetPro format.
multimodal-data	Combined video (npy) and tracking (parquet) data with 1 file per event per modality. Uses a unified annotation file for both modalities in SoccerNetPro format.

Benchmark Results

Pixels vs. Positions

Modality	Model	Params	Balanced Acc.	Training Time
Tracking	GIN + Attention + Positional Edges	197K	67.2%	4 GPU hours
Video	VideoMAEv2 (finetuned)	86.3M	58.1%	34 GPU hours

The tracking model outperforms the video baseline by 9.1 percentage points while using 438x fewer parameters and training 4.25x faster.

Per-Class Comparison (Test Set)

Class	Samples	Tracking	Video
PASS	9,255	70.1	65.2
TACKLE	1,697	50.7	57.8
OUT	955	85.7	78.7
HEADER	872	75.0	55.1
HIGH PASS	405	27.2	40.5
THROW IN	393	65.6	40.5
CROSS	373	81.8	72.9
FREE KICK	273	86.5	71.1
SHOT	266	73.2	58.7
GOAL	30	56.7	36.7
Overall	14,519	67.2	57.7

Tracking excels on spatially distinctive events (OUT, FREE KICK, CROSS, SHOT, GOAL), while video outperforms on HIGH PASS (+13.3%) and TACKLE (+7.1%) where visual cues like ball trajectory and body dynamics provide discriminative information.

Uses

Direct Use

• Benchmarking video-based vs. tracking-based group activity recognition
• Training and evaluating GAR models on football broadcast data
• Studying multimodal fusion approaches combining visual and positional features
• Analyzing spatial interaction patterns in team sports

Dataset Creation

Curation Rationale

No standardized benchmark previously existed that aligns broadcast video and tracking data for the same group activities. This made fair, apples-to-apples comparison between video-based and tracking-based approaches impossible. SoccerNet-GAR was created to fill this gap by providing synchronized multimodal observations under a unified evaluation protocol.

Source Data

The dataset was constructed from the PFF FC website (now Gradient Sports), which provides comprehensive broadcast videos, player tracking data, and event annotations across all 64 FIFA World Cup 2022 tournament matches.

Annotation Process

Event annotations with precise timestamps were created by trained annotators and verified through quality control procedures by PFF FC using both video and tracking views. Each event is labeled with one of 10 group activities and temporally marked at the moment of occurrence.

For each annotated event at timestamp t_e, a 4.5-second temporal window centered at t_e is extracted from both broadcast video and player tracking streams, with 16 samples taken at 30 fps with a 9-frame interval.

Comparison with Existing Datasets

Dataset	Year	Domain	Events	Classes	Modalities
CAD	2009	Pedestrian	2,511	5	V
Volleyball	2016	Volleyball	4,830	8	V
SoccerNet	2018	Football	6,637	3	V
NBA	2020	Basketball	9,172	9	V
SoccerNet-v2	2021	Football	110,458	17	V
SoccerNet-BAS	2024	Football	11,041	12	V
FIFAWC	2024	Football	5,196	12	V
SoccerNet-GAR	2025	Football	94,285	10	V + T

SoccerNet-GAR is the second largest GAR dataset (after SoccerNet-v2) and the only one providing synchronized video and tracking modalities.

Citation

@article{karki2025pixels,
  title={Pixels or Positions? Benchmarking Modalities in Group Activity Recognition},
  author={Karki, Drishya and Ramazanova, Merey and Cioppa, Anthony and Giancola, Silvio and Ghanem, Bernard},
  journal={arXiv preprint arXiv:2511.12606},
  year={2025}
}

Authors

• Drishya Karki (KAUST)
• Merey Ramazanova (KAUST)
• Anthony Cioppa (University of Liege)
• Silvio Giancola (KAUST)
• Bernard Ghanem (KAUST)

Contact

• drishya.karki@kaust.edu.sa / karkidrishya1@gmail.com
• silvio.giancola@kaust.edu.sa