Detecto - Real-Time Object Detection Model

Detecto is an AI-powered object detection system that identifies and locates multiple objects in real-time video streams. It uses state-of-the-art deep learning models to deliver fast, accurate, and reliable object detection while ignoring person detections for privacy purposes.

Features

• 🚀 Real-time object detection using DETR (DEtection TRansformer) model
• 🔒 Privacy-focused: Automatically ignores person detections
• 🎥 Live camera feed processing with bounding boxes and labels
• ⚡ Optimized for performance with multi-threading
• 🖼️ Image and webcam snapshot testing capabilities
• 🧠 91 object classes detection (excluding persons)
• 🖥️ Cross-platform compatibility

Installation

1. Clone the repository:

   git clone https://github.com/CrazyBong/Detecto-Object-detection-model.git
   cd Detecto-Object-detection-model

2. Install the required dependencies:

   pip install -r requirements.txt

3. Run the application:

   python main.py

Usage

Real-time Camera Detection

Run the main application to start real-time object detection:

python main.py

Controls:

• Press 'q' to quit the application

Testing Suite

The project includes a comprehensive testing suite:

python test.py

Available test modes:

1. Test on existing image files
2. Capture from webcam and test detection
3. Quick model verification

Technical Details

Model

The system uses the facebook/detr-resnet-50 model from Hugging Face Transformers, which is based on the DETR (DEtection TRansformer) architecture. This model can detect 91 different object classes with high accuracy.

Performance Optimizations

• Multi-threading: Separates camera capture, object detection, and display into different threads
• Frame skipping: Processes every Nth frame to maintain real-time performance
• GPU acceleration: Automatically utilizes CUDA if available
• Confidence thresholding: Filters low-confidence detections

Privacy Features

• Person detection filtering: All person detections are automatically filtered out
• No data storage: No images or videos are saved by default
• Local processing: All processing happens on your device

Requirements

See requirements.txt for detailed dependencies.

Project Structure

.
├── main.py              # Real-time camera object detection
├── test.py              # Image and webcam testing suite
├── requirements.txt     # Python dependencies
├── README.md            # This file
└── .gitignore           # Git ignore rules

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

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

Acknowledgments

• DETR model by Facebook AI Research
• Hugging Face Transformers library
• OpenCV for computer vision operations