An advanced drone detection system combining a Next.js web interface, deep learning models, and synthetic dataset generation using Blender. This project is designed to detect UAVs (Unmanned Aerial Vehicles) in real-time, offering high accuracy and adaptability across various scenarios.
The UAV Detection System is a comprehensive solution for detecting rones. It integrates state-of-the-art computer vision models with a user-friendly web interface, providing a versatile tool for drone surveillance in diverse environments.
Detection Modes:
- Photo analysis
- Video processing
- Real-time camera stream detection
-
Data Pipeline:
- Synthetic data generation with Blender
- Augmentation for enhanced model robustness
-
Model:
- YOLO-based architectures fine-tuned for UAV detection
-
Web Interface:
- Next.js for real-time visualization and user interaction
The dataset creation process employs Blender for rendering diverse and realistic UAV scenarios:
| Category | Details |
|---|---|
| Scene Setup | 3 UAV models, 50 unique environments, 25 flight paths, variable lighting and weather conditions |
| Generation Parameters | Resolution: 1920x1080, Classes: 3, Total Images: 2,331 |
The model training pipeline uses a YOLOv10-based architecture optimized for drone detection. Metrics, checkpoints, and configurations are saved for reproducibility.
| Metric | Performance on Test Data | Performance on Validation Data | Difference in Performance |
|---|---|---|---|
| mAP50 | 0.9825 | 0.98966 | -0.00716 |
| mAP50-95 | 0.6912 | 0.75729 | -0.06609 |
Ground Truth vs Test Images
|
Ground Truth vs Real Images
|
Contributions are welcome! Please fork the repository, create a branch, and submit a pull request.
- Fork the repository
- Create a feature branch (
git checkout -b feature-name) - Commit your changes (
git commit -m 'Add feature') - Push to the branch (
git push origin feature-name) - Open a pull request
This project is licensed under the MIT License. See the LICENSE file for more details.