datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
SoccerNet_raw_HQ
SoccerNet Raw HQ
SoccerNet Raw HQ provides controlled access to the original SoccerNet broadcast videos and several research-ready derivatives. The repository is intended for academic research in soccer video understanding, including action spotting, temporal localization, representation learning, video retrieval, and multimodal analysis.
The original SoccerNet release contains 500 complete broadcast games, covering six major European competitions, three seasons from 2014 to… See the full description on the dataset page: https://huggingface.co/datasets/SoccerNet/SoccerNet_raw_HQ.SoccerHigh
⚽ SoccerHigh
This dataset provides annotations and pre-extracted features for the SoccerHigh benchmark introduced in:
SoccerHigh: A Benchmark Dataset for Automatic Soccer Video Summarization
Artur Díaz-Juan, Coloma Ballester, Gloria HaroACM MMSports 2025
📦 Contents
Highlight summary annotations
Train / validation / test splits
Pre-extracted visual features (no raw videos)
All data are provided as .npy feature arrays, .srt temporal annotations, and .json metadata files.… See the full description on the dataset page: https://huggingface.co/datasets/imva-upf/SoccerHigh.soccer-dataset
Global Football (Soccer) Data Lake
Cleaned, deduplicated, quality-gated football match data for BTTS / goals modelling.
Sources: API-Football + football-data.co.uk. Pipeline & docs:
https://github.com/eatpizzanot/soccer-dataset
673,966 fixtures (644,901 played), 271 leagues,
11,104 teams, 2008-06-07 - 2027-06-06.
BTTS base rate 0.5063. xG fake-zeros removed; known_at leakage guard;
12-dimension QA gate (QUALITY_REPORT.md).
Caveats: league history is uneven — check… See the full description on the dataset page: https://huggingface.co/datasets/eatpizzanot/soccer-dataset.SoccerNet-V3
Dataset Card for SoccerNet-V3
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.
This is a… See the full description on the dataset page: https://huggingface.co/datasets/Voxel51/SoccerNet-V3.SN-LabelsSN-Features
SoccerNet Features
Pre-extracted per-game features for the SoccerNet benchmark, structured as <league>/<season>/<game>/<file>, one file per game half (1_.../2_...).
This main branch holds no data — each feature type lives on its own branch so you only download what you need:
Branch
Files
Description
baidu-soccer-embeddings
{1,2}_baidu_soccer_embeddings.npy
Frame embeddings from baidu-research/vidpress-sports, used by the Action Spotting and Dense Video Captioning 2023… See the full description on the dataset page: https://huggingface.co/datasets/SoccerNet/SN-Features.SoccerWiki
Dataset Card for SoccerWiki
This repository contains the database for paper "Multi-Agent System for Comprehensive Soccer Understanding" in ACM Multimidia 2025.
SoccerWiki is a large-scale multimodal soccer knowledge base. The dataset integrates rich domain knowledge about soccer players, teams, referees, and venues, which is used to facilitate knowledge-driven reasoning and decision-making in various soccer-related tasks.
This dataset was built using data from Wikipedia and… See the full description on the dataset page: https://huggingface.co/datasets/SJTU-AI4Sports/SoccerWiki.soccertrack-v2
SoccerTrack v2
Ten university-level soccer matches (934 minutes) recorded by fixed panoramic
camera systems whose field of view spans the entire pitch, released together
with frame-level game state annotations and player-linked ball action events
on the same footage.
Contents
Folder
Contents
videos/
20 panoramic half-match videos (two 45-minute periods per match)
gsr/
Game state annotations: one JSON per half (pitch coordinates, jersey numbers and… See the full description on the dataset page: https://huggingface.co/datasets/atomscott/soccertrack-v2.SN-GSR-2025
SoccerNet Challenge 2025 - Game State Reconstruction
Download the dataset
Install the huggingface_hub pip package:
pip install huggingface_hub[cli]
Download the dataset with the following Python code :
from huggingface_hub import snapshot_download
snapshot_download(repo_id="SoccerNet/SN-GSR-2025",
repo_type="dataset", revision="main",
local_dir="SoccerNet/SN-GSR-2025")
SN-GSR-2024SN-Tracking-2023soccer-dialoguesSN-PCBAS-2026SN-BAS-2025
SoccerNet Challenge 2025 - Team Ball Action Spotting
Download the dataset
Install the huggingface_hub pip package:
pip install huggingface_hub[cli]
Download the dataset with the following Python code :
from huggingface_hub import snapshot_download
snapshot_download(repo_id="SoccerNet/SN-BAS-2025",
repo_type="dataset", revision="main",
local_dir="SoccerNet/SN-BAS-2025")
Unzip the dataset splits
The zipped folder contains… See the full description on the dataset page: https://huggingface.co/datasets/SoccerNet/SN-BAS-2025.SoccerNet_Field_SegmentationProcessed data from the Soccernet 2023 dataset. Processing notebook is included in this repo.
To see an example:
def show_item(item):
fig, axs = plt.subplots(nrows = 1, ncols = 4, figsize = (20, 4))
axs[0].imshow(item['image'])
axs[0].set_title("Image")
axs[0].axis('off')
axs[1].imshow(overlay_mask(item['image'], item['outlines']))
axs[1].set_title("Outlines")
axs[1].axis('off')
axs[2].imshow(show_segments(item['segments']))
axs[2].set_title("Segments")… See the full description on the dataset page: https://huggingface.co/datasets/nreHieW/SoccerNet_Field_Segmentation.soccer-events
Soccer (Football) In-Match Events
A curated sample dataset of in-match soccer/football events, automatically identified using multi-modal models from Infactory. This dataset sample contains video clips and metadata for three key event types: yellow cards, red cards, and goals. Infactory models are capable of detecting corner kicks, blocked shots, saved shots, substitutions, and many more types of events in gameplay footage.
Dataset Description
This dataset provides… See the full description on the dataset page: https://huggingface.co/datasets/infactory-ai/soccer-events.soccernetpro-localization-tennis
SoccerNetPro Localization (Tennis)
This repository provides a tennis action spotting / localization dataset in an OpenSportsLab / SoccerNet-style format.
The dataset is organized by split (train/, valid/, test/) with video clips and corresponding localization annotations in JSON.
📌 Task
Task type: action_spotting (a.k.a. temporal action localization / event spotting)
Annotation granularity: clip-relative timestamps in milliseconds (position_ms)
Label format:… See the full description on the dataset page: https://huggingface.co/datasets/OpenSportsLab/soccernetpro-localization-tennis.SoccerNet-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 87,939 annotated group activities across 10 action classes, enabling direct comparison between video-based and tracking-based approaches.
Dataset Details… See the full description on the dataset page: https://huggingface.co/datasets/OpenSportsLab/SoccerNet-GAR.SN-MVFouls-2025
SoccerNet Challenge 2025 - Multi-View Fouls Recognition
Download the dataset
Install the huggingface_hub pip package:
pip install huggingface_hub[cli]
Download the dataset with the following Python code :
from huggingface_hub import snapshot_download
snapshot_download(repo_id="SoccerNet/SN-MVFouls-2025",
repo_type="dataset", revision="main",
local_dir="SoccerNet/SN-MVFouls-2025")
Unzip the dataset splits
The zipped… See the full description on the dataset page: https://huggingface.co/datasets/SoccerNet/SN-MVFouls-2025.SN-VQA-2026
2026 Soccernet Challenge - VQA Overview
Task
Soccernet-VQA is a challenge focused on multimodal (text, image, video) multiple-choice question answering, covering 14 distinct soccer understanding tasks. These tasks include assessing background knowledge of players and teams, determining camera status, classifying actions, recognizing fouls, and many other complex scenarios.
More details could be found at:
🏆 2026 SoccerNet Challenge - VQA: DevKit ⋅ Rules ⋅ Eval (Test) ⋅… See the full description on the dataset page: https://huggingface.co/datasets/SoccerNet/SN-VQA-2026.kaggle-hugomathien-soccerSource: https://www.kaggle.com/datasets/hugomathien/soccer by Hugo Mathien
About Dataset
The ultimate Soccer database for data analysis and machine learning
What you get:
+25,000 matches
+10,000 players
11 European Countries with their lead championship
Seasons 2008 to 2016
Players and Teams' attributes* sourced from EA Sports' FIFA video game series, including the weekly updates
Team line up with squad formation (X, Y coordinates)
Betting odds from up to 10 providers… See the full description on the dataset page: https://huggingface.co/datasets/julien-c/kaggle-hugomathien-soccer.SN-Calibration-2023robocolosseum-so101-soccer-red_bowl-40episodesThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v3.0",
"fps": 30,
"features": {
"action": {
"dtype": "float32",
"names": [
"shoulder_pan.pos",
"shoulder_lift.pos",
"elbow_flex.pos",
"wrist_flex.pos",
"wrist_roll.pos",
"gripper.pos"
],
"shape": [
6… See the full description on the dataset page: https://huggingface.co/datasets/tsangb34/robocolosseum-so101-soccer-red_bowl-40episodes.soccernet-500-videosSN-MVFouls-2024soccer_ball_raw_tactileThis dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.1",
"robot_type": "Unitree_G1_Inspire",
"total_episodes": 81,
"total_frames": 29879,
"total_tasks": 1,
"total_videos": 162,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 30,
"splits": {
"train": "0:81"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/eunjuri/soccer_ball_raw_tactile.SN-echoes[Paper] | [GitHub]
Dataset Card for SoccerNet-Echoes
This dataset card aims to provide comprehensive details for the SoccerNet-Echoes dataset, an audio commentary dataset for soccer games.
Dataset Details
Dataset Description
SoccerNet-Echoes is an audio commentary dataset for soccer games, curated by SimulaMet under the AI-Storyteller project. It is funded by the Research Council of Norway (project number 346671) and shared by the SoccerNet team. The dataset… See the full description on the dataset page: https://huggingface.co/datasets/SoccerNet/SN-echoes.ActionAnticipation
SoccerNet Challenge 2026 - Action Anticipation Dataset
The SoccerNet Action Anticipation dataset splits the 2024 ball action spotting dataset into 30 second clips, which can then be used to anticipate between 10 action classes that will happen 5 seconds into the future. This is the dataset used for 2026 SoccerNet Action Anticipation challenge.
Relevant Links
Repository: https://github.com/MohamadDalal/FAANTRA
Paper: https://huggingface.co/papers/2504.12021… See the full description on the dataset page: https://huggingface.co/datasets/SoccerNet/ActionAnticipation.SN-BAS-2024OSL-SoccerNet