HockeyAI: A Multi-Class Ice Hockey Dataset for Object Detection

🏒 Overview

The HockeyAI dataset is an open-source dataset designed to advance computer vision research in ice hockey. With 2,100 high-resolution frames and detailed YOLO-format annotations, this dataset provides a foundation for object detection in fast-paced sports environments. It is ideal for researchers and developers working on player and puck tracking, event detection, and real-time AI-driven analytics.

📌 Classes

The dataset includes annotations for seven key classes:

• centerIce: Center circle on the rink
• faceoff: Faceoff dots
• goal: Goal frame
• goaltender: Goalkeeper
• player: Ice hockey players
• puck: The small, fast-moving object central to gameplay
• referee: Game officials

📥 Access the Dataset

The dataset is hosted on Hugging Face:

🔗 Dataset Path: HockeyAI on Hugging Face

🎯 Pretrained Model

We provide a trained object detection model, available on Hugging Face.

🔗 Trained Model: HockeyAI Model

🚀 Live Demo

Explore the HockeyAI object detection demo on Hugging Face Spaces.

🔗 HockeyAI Demo: Hugging Face Space

🤝 Citation

If you use this dataset or model in your research, please cite:

@inproceedings{HockeyAI,
    title = {{HockeyAI: A Multi-Class Ice Hockey Dataset for Object Detection}},
    author = {Houshmand Sarkhoosh, Mehdi and Gautam, Sushant and Midoglu, Cise and Sabet, Saeed Shafiee and Kupka, Tomas and Halvorsen, Pål},
    publisher = {Association for Computing Machinery},
    booktitle = {Proceedings of the 16th ACM Multimedia Systems Conference},
    address = {New York, NY, USA},
    url = {https://doi.org/10.1145/3712676.3718335},
    isbn = {9798400714672},
    location = {Stellenbosch, South Africa},
    year = {2025},
    doi={10.1145/3712676.3718335}
}

📩 Contact

For inquiries, collaborations, or questions:

• 📧 Mehdi Houshmand: mehdi@forzasys.com
• 📧 Cise Midoglu: cise@forzasys.com
• 📧 Pål Halvorsen: paalh@simula.no

---

⚡ HockeyAI is part of ongoing AI-driven sports analytics research.