Blue Iris Custom AI Models

Custom trained AI models optimized for use with Blue Iris video surveillance software.

Overview

This repository contains custom ONNX models trained specifically for detection tasks in Blue Iris. These models are optimized for accuracy and performance in video surveillance scenarios.

Models Included

License Plate Detection Model

• Path: custom-license-plates-model/plates.onnx
• Framework: YOLO11n (nano variant)
• Format: ONNX (Open Neural Network Exchange)
• Task: Object Detection
• Classes: 1 (license_plate)

Requirements

• Blue Iris version 6.0.1.2 or higher with AI support
• Compatible AI processing backend (CPU or GPU with DirectML support)

Installation

1. Clone this repository or download the model files
2. Copy the desired .onnx model file to your Blue Iris AI models directory

Model Performance

The license plate detection model achieves:

• Precision: 98.9%
• Recall: 97.3%
• mAP50: 99.3%
• mAP50-95: 85.9%

Trained on 42,024 training images and validated on 11,988 validation images over 300 epochs.

Model Details

License Plate Model Specifications

• Architecture: YOLO11n (lightweight nano variant)
• Input Size: 640×640 pixels
• Parameters: 2,590,035
• Training Device: NVIDIA GeForce RTX 5090
• Training Framework: Ultralytics YOLO 8.3.195
• Inference Format: TorchScript ONNX

Training Configuration

• Epochs: 300
• Batch Size: 128
• Optimizer: SGD with automatic tuning
• Image Size: 640×640
• Augmentation: Mosaic, mixup, HSV adjustments, flip, scale, translate

Repository Structure

Blue-Iris-Custom-AI-Models/
├── LICENSE                           # MIT License
├── README.md                         # This file
└── custom-license-plates-model/      # License plate detection model
    ├── plates.onnx                   # ONNX model file
    └── plates.onnx_training_results/ # Training metrics and logs
        ├── args.yaml                 # Training arguments
        ├── results.csv               # Training results data
        └── training-results.txt      # Complete training log

License

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

Contributing

Contributions are welcome! If you have improvements or additional models to share:

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Submit a pull request

Support

For issues or questions:

• Open an issue in this repository
• Consult Blue Iris documentation for integration help
• Check the model training results for performance metrics

Acknowledgments

• Built with Ultralytics YOLO
• Designed for Blue Iris video surveillance software
• Trained using PyTorch and exported to ONNX format

Version History

• Initial Release: License plate detection model based on YOLO11n
  • High accuracy detection (99.3% mAP50)
  • Optimized for real-time inference
  • Lightweight architecture suitable for various hardware configurations

Technical Notes

Model Export

The models are exported using TorchScript format with simplification enabled for optimal inference performance.

Hardware Requirements

• Minimum: Modern CPU with AVX2 support
• Recommended: NVIDIA GPU with DirectML support for real-time processing
• Memory: At least 4GB RAM for model loading and inference

Performance Tips

• Use GPU acceleration when available
• Adjust confidence thresholds based on your environment
• Configure appropriate detection zones to reduce false positives
• Monitor system resources during high camera count deployments