⚽ Football Image Dataset

A manually annotated object-detection dataset of football (soccer) match frames, labelled in YOLO format with three classes: ball, player and referee.

---

📋 Overview

This dataset contains 612 images (Full HD, 1920×1080) extracted from football match footage, each one manually annotated with bounding boxes. It is ready to train modern object detectors such as the Ultralytics YOLO family.

Class id	Name	Description
0	ball	The match ball
1	player	Outfield players and goalkeepers
2	referee	Referee and assistant referees

📊 Dataset statistics

Split	Images	ball	player	referee	Total boxes
train	512	417	7,278	634	8,329
val	100	83	1,566	158	1,807
Total	612	500	8,844	792	10,136

The dataset is split into an 84 / 16 train–validation partition (512 / 100 images). One training image is a background sample (no objects), which is intentional and helps reduce false positives.

📁 Repository structure

football-image-dataset/
├── images/
│   ├── train/        # 512 .jpg frames
│   └── val/          # 100 .jpg frames
├── labels/
│   ├── train/        # 512 .txt YOLO annotations
│   └── val/          # 100 .txt YOLO annotations
├── data.yaml         # Ultralytics dataset configuration
├── LICENSE           # CC BY 4.0
└── README.md

🏷️ Annotation format

Labels follow the YOLO convention: one .txt file per image, with the same base name. Each line describes one bounding box:

<class_id> <x_center> <y_center> <width> <height>

All coordinates are normalised to [0, 1] relative to the image dimensions. Example (labels/train/<name>.txt):

1 0.387760 0.541667 0.027604 0.105556   # player
2 0.353385 0.535648 0.019271 0.108333   # referee
0 0.294531 0.541667 0.006771 0.011111   # ball

📜 License

Released under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. You are free to use, share and adapt the dataset — including commercially — as long as you give appropriate credit.

📣 Citation

If you use this dataset in your work, please cite it:

@misc{football_image_dataset,
  title        = {Football Image Dataset: YOLO-annotated ball, player and referee detection},
  author       = {skaczylo},
  year         = {2026},
  howpublished = {\url{https://github.com/skaczylo/football-image-dataset}}
}

⭐ Support

If you find this dataset useful, please consider giving it a star on GitHub — it helps others discover the project and is greatly appreciated!

---

_{Keywords: football dataset · soccer dataset · YOLO dataset · object detection · ball detection · player detection · referee detection · sports analytics · computer vision · annotated images · bounding boxes · YOLOv8 · YOLO11 · Ultralytics · deep learning · machine learning · image dataset · training data}