Cephalometry Using YOLOv9

A computer vision project that applies YOLOv9 object detection for automated cephalometric analysis in dental and orthodontic imaging. This project enables precise detection and measurement of anatomical landmarks in cephalometric X-ray images.

📋 Table of Contents

• Overview
• Features
• Requirements
• Installation
• Project Structure
• Usage
• Model Configuration
• Detection Parameters
• Output Formats
• Contributing
• License

🔍 Overview

Cephalometry is a diagnostic tool used in orthodontics and oral surgery to analyze the relationships between dental and skeletal structures. This project leverages the power of YOLOv9, a state-of-the-art object detection model, to automatically identify and locate anatomical landmarks in cephalometric radiographs.

Key Benefits:

• Automated Analysis: Reduces manual measurement time and human error
• High Precision: YOLOv9's advanced architecture ensures accurate landmark detection
• Clinical Integration: Designed for integration into dental practice workflows
• Flexible Output: Multiple output formats for different use cases

✨ Features

• Real-time Detection: Fast inference on cephalometric images
• Multiple Input Sources: Support for images, videos, webcam, and URLs
• Customizable Confidence Thresholds: Adjustable detection sensitivity
• Multiple Output Formats:
  • Annotated images with bounding boxes
  • Text files with coordinates
  • Cropped landmark regions
• GPU Acceleration: CUDA support for faster processing
• Batch Processing: Process multiple images simultaneously

🛠️ Requirements

System Requirements

• Python 3.8 or higher
• CUDA-compatible GPU (recommended for faster inference)
• Minimum 8GB RAM
• OpenCV-compatible system

Core Dependencies

torch>=1.8.0
torchvision>=0.9.0
opencv-python>=4.5.0
numpy>=1.21.0
Pillow>=8.0.0
PyYAML>=5.4.0
tqdm>=4.60.0
matplotlib>=3.3.0
seaborn>=0.11.0
pandas>=1.3.0
pathlib
argparse

YOLOv9 Specific Libraries

ultralytics
thop
tensorboard
protobuf<4.21.3

📦 Installation

1. Clone the Repository

git clone https://github.com/selvatharrun/cephalometry-using-yolov9-.git
cd cephalometry-using-yolov9-

2. Create Virtual Environment (Recommended)

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

3. Install Dependencies

pip install -r requirements.txt

4. Install PyTorch with CUDA Support (if available)

# For CUDA 11.8
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118

# For CPU only
pip install torch torchvision torchaudio

5. Download Pre-trained Models

# Download YOLOv9 weights (replace with your trained cephalometry model)
wget https://github.com/WongKinYiu/yolov9/releases/download/v0.1/yolov9c.pt

📁 Project Structure

cephalometry-using-yolov9-/
├── detect.py                 # Main detection script
├── models/
│   ├── common.py            # Model architecture components
│   └── ...
├── utils/
│   ├── dataloaders.py       # Data loading utilities
│   ├── general.py           # General utility functions
│   ├── plots.py             # Visualization utilities
│   └── torch_utils.py       # PyTorch utilities
├── data/
│   ├── images/              # Input images directory
│   └── coco.yaml            # Dataset configuration
├── runs/
│   └── detect/              # Output directory
├── requirements.txt         # Python dependencies
└── README.md               # Project documentation

🚀 Usage

Basic Detection

python detect.py --source data/images --weights yolo.pt

Advanced Usage Examples

Process Single Image

python detect.py --source path/to/image.jpg --weights yolo.pt --conf-thres 0.5

Process Video

python detect.py --source path/to/video.mp4 --weights yolo.pt --save-txt

Webcam Detection

python detect.py --source 0 --weights yolo.pt --view-img

Batch Processing with Custom Output

python detect.py \
    --source data/images \
    --weights yolo.pt \
    --project runs/cephalometry \
    --name experiment1 \
    --save-txt \
    --save-conf \
    --conf-thres 0.6 \
    --iou-thres 0.4

⚙️ Model Configuration

Dataset Configuration (data/coco.yaml)

# Modify for cephalometric landmarks
names:
  0: nasion
  1: sella
  2: orbitale
  3: porion
  4: anterior_nasal_spine
  5: posterior_nasal_spine
  6: pogonion
  7: menton
  8: gonion
  9: articulare

Detection Parameters

Parameter	Description	Default	Recommended Range
--conf-thres	Confidence threshold	0.25	0.3-0.7
--iou-thres	IoU threshold for NMS	0.45	0.4-0.6
--imgsz	Input image size	640	640-1280
--max-det	Maximum detections	1000	50-200
--line-thickness	Bounding box thickness	1	1-5

📊 Output Formats

1. Annotated Images

• Images with bounding boxes around detected landmarks
• Confidence scores displayed
• Saved in runs/detect/exp/

2. Text Files (--save-txt)

Format: class x_center y_center width height confidence

0 0.5234 0.3456 0.0234 0.0345 0.89
1 0.6123 0.4567 0.0198 0.0276 0.92

3. Cropped Images (--save-crop)

• Individual landmark regions saved as separate images
• Useful for detailed analysis

🔧 Customization

Training Your Own Model

1. Prepare cephalometric dataset with landmark annotations
2. Configure dataset YAML file
3. Train using YOLOv9 training script
4. Replace yolo.pt with your trained weights

Adjusting for Clinical Use

• Modify confidence thresholds based on clinical requirements
• Customize landmark classes in configuration
• Implement post-processing for measurements

🤝 Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/improvement)
3. Commit changes (git commit -am 'Add new feature')
4. Push to branch (git push origin feature/improvement)
5. Create Pull Request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• YOLOv9 team for the excellent object detection framework
• OpenCV community for computer vision tools
• PyTorch team for the deep learning framework

---

For more detailed information about YOLOv9 architecture and training procedures, refer to the original YOLOv9 documentation.