This project provides an AI-powered basketball referee system designed to detect travel and double dribble violations in real-time. Targeting referees, coaches, and players, it aims to enhance game fairness and provide valuable data for analysis by leveraging advanced computer vision techniques.

How It Works

The system employs a custom YOLO model for real-time basketball detection and YOLO pose estimation to identify key player body joints. By analyzing the spatial and temporal relationships between detected basketballs and player keypoints, it applies predefined rules to identify travel and double dribble infractions, offering visual feedback on the video feed.

Quick Start & Requirements

• Install via pip install ultralytics.
• Requires Python 3.11.
• The core basketballModel.pt file must be downloaded separately from a provided Google Drive link due to GitHub storage limits.
• Official documentation and demo videos are available via provided links.

Highlighted Details

• Real-time detection of travel and double dribble violations.
• Utilizes YOLO object detection and pose estimation.
• Customizable parameters for detection sensitivity.
• Potential for expansion to detect other basketball violations.

Maintenance & Community

The project has received media attention from sources like Georgia Tech, Hackster.io, and AIFinityHub, indicating external interest and validation. Links to social media and webinars are provided for community engagement and further technical details.

Licensing & Compatibility

The repository does not explicitly state a license. Users should verify licensing terms for commercial use or integration into closed-source projects.

Limitations & Caveats

The project relies on a large, externally hosted model file (basketballModel.pt) which is a potential point of failure or inconvenience for users. The accuracy and robustness of the detection algorithms in diverse real-world game scenarios are not detailed.