A robust real-time AI system leveraging CCTV and mobile video streams to detect, classify, and estimate waste in public areas using deep learning, with geospatial mapping for smart waste management.
Vision2Clean AI is a solution for the subject "Waste to Wealth to Wheels".
It utilizes YOLOv11 and advanced computer vision to:
- Detect and classify waste types in real-time from mobile/CCTV feeds
- Estimate waste quantity via bounding box analysis
- Tag detections with static GPS coordinates (pre-mapped per camera)
- Visualize results on a live interactive map
- 🧠 YOLOv11-powered multi-category waste detection
- 🗂️ Supports detection of Plastic, Organic, Paper, Metal, E-waste, and more
- 🛰️ Location tagging using static GPS data
- 🗺️ Interactive map visualization with Folium
- 💾 Data logging to CSV and SQLite database
- 📲 Mobile phone camera integration for live feeds
- 📊 Dashboard for analytics (React.js)
| Component | Technology/Library |
|---|---|
| Detection Model | YOLOv11 (Ultralytics) |
| Video Streaming | IP Webcam App / OpenCV |
| Mapping | Folium (Python) |
| Data Storage | CSV / SQLite |
| Dashboard | React.js |
git clone https://github.com/yourusername/vision2clean-ai.git
cd vision2clean-ai# Python dependencies
pip install -r requirements.txt
# For dashboard (React.js)
cd dashboard
npm install- Set up IP Webcam App or connect CCTV feed.
- Update camera GPS coordinates in
config/cameras.json.
# Start detection and mapping
python main.py
# Start dashboard (in dashboard directory)
npm startvision2clean-ai/
├── config/
│ └── cameras.json
├── data/
│ └── detections.csv
├── dashboard/
├── main.py
├── requirements.txt
└── README.md
Contributions are welcome! Please open issues or submit pull requests for improvements.
This project is licensed under the MIT License.