HockeyAI YOLOv8 Model

🔗 This model is trained on the HockeyAI dataset.

• 📊 Access the dataset used for training here: https://huggingface.co/datasets/SimulaMet-HOST/HockeyAI
• 🚀 Try the model in action with our interactive Hugging Face Space: https://huggingface.co/spaces/SimulaMet-HOST/HockeyAI

Model Overview

The HockeyAI project provides a YOLOv8 medium model fine-tuned on the HockeyAI dataset. This model serves as a benchmark for ice hockey object detection tasks and achieves high performance across all seven classes defined in the dataset.

Model Performance

The model was evaluated on a holdout set of the HockeyAI dataset, achieving the following performance metrics:

• Mean Average Precision (mAP@0.5): XX.X%
• Precision: 100% for all classes
• Recall: 95% for all classes
• F1-Score: 93% for all classes

Usage

The pretrained model is available in this repository as a .pt file. You can download and use it directly with the YOLOv8 framework for:

• Inference on new hockey videos or images
• Further fine-tuning on your specific use case
• Benchmarking against new approaches

Supported Classes

The model is trained to detect seven classes:

• Center Ice
• Faceoff Dots
• Goal Frame
• Goaltender
• Players
• Puck
• Referee

Requirements

• YOLOv8 framework
• Python 3.7+
• PyTorch 1.7+

Getting Started

1. Download the model weights from this repository
2. Install the required dependencies
3. Load and use the model with YOLOv8's standard API

📩 For any questions regarding this project, or to discuss potential collaboration and joint research opportunities, please contact:

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