Advanced Pickleball Analysis

A comprehensive desktop application designed for automated pickleball match analysis using computer vision and deep learning.

Features

• Object Detection: Real-time tracking of players, paddles, and the pickleball using YOLOv8/v11.
• Court Mapping: Automated court keypoint detection and homography estimation for top-down analytics.
• Live Analysis: Integrated GUI for real-time video processing and visualization.
• Analytics & Insights:
  • Trajectory interpolation for missing ball detections.
  • Interactive heatmaps of player and ball movements.
  • Event logging and session summaries.
• YouTube Support: Directly analyze matches from YouTube URLs using yt-dlp.
• Data Export: Export results to MP4 (annotated video), CSV (event log), and JSON (summary).

Repository Structure

.
├── data/               # Sample videos and test data
├── models/             # Pre-trained YOLO models (.pt)
├── outputs/            # Generated analysis reports and videos
├── pickleball_analysis/ # Core package source code
│   ├── core/           # Analysis logic, models, and tracking
│   └── gui/            # CustomTkinter interface components
├── main.py             # Application entry point
├── requirements.txt    # Project dependencies
└── README.md           # Project documentation

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/pickleball-analysis.git
   cd pickleball-analysis

2. Create a virtual environment (recommended):

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

   Note: Ensure you have ffmpeg installed on your system for video processing.

Usage

Run the main application using the following command:

python main.py

CLI Arguments

You can also customize the launch parameters via command line:

python main.py --source data/test.mp4 --device 0 --imgsz 960 --heatmap

• --source: Path to a video file or YouTube URL.
• --device: Torch device (e.g., 0, 1, or cpu).
• --heatmap: Enable heatmap generation by default.
• --court-model: Path to a custom court pose model.
• --object-model: Path to a custom object detection model.

Core Dependencies

• Ultralytics: Powering the YOLO object detection and pose models.
• CustomTkinter: Providing a modern desktop UI.
• OpenCV: For high-performance video frame processing.
• yt-dlp: For seamless YouTube video integration.

Status & Contributing

⚠️ Note: This project is still in early development and may contain bugs.

Contributions are welcome! If you encounter issues or have ideas for improvements, please feel free to:

• Open an Issue to report a bug.
• Submit a Pull Request (PR) with your fixes or enhancements.

License

Distributed under the MIT License. See LICENSE for more information.