Illegal Site Detection Tool

This Streamlit web app predicts whether a given GPS location is a legal or illegal dumping site using:

• A classification model trained on spatial and visual features
• YOLOv7-based object detection for garbage and buildings
• Real satellite tile images fetched from map servers

Features

• Select a location by entering coordinates or pinning a point on the map
• Fetch satellite tile of the selected location
• Extract spatial features (buildings, roads, police proximity, etc.) using OpenStreetMap
• Classify site legality using a trained model (classModel.h5)
• Run object detection using YOLOv7 weights (garbage.pt, building.pt)
• View prediction results and detected bounding boxes directly on the app

How to Run

1. Clone the Repository

git clone https://github.com/your-username/your-repo-name.git
cd your-repo-name

2. Install Dependencies

pip install -r requirements.txt

3. Add Models

Put your trained model files in the model_wts/ directory:

• garbage.pt
• building.pt
• classModel.h5

4. Launch the App

streamlit run app.py

Requirements

See requirements.txt for the full list. Major ones include:

• streamlit
• torch
• tensorflow
• opencv-python
• osmnx
• shapely
• folium
• geopy

Notes

• The app fetches satellite tiles from public map tile servers.
• It also queries OpenStreetMap for nearby features.
• Internet connection is required for full functionality.

Output Example

• Tile image with object detection:

  • Detected buildings
  • Detected garbage

• Final legality prediction (with score)

Credits

Created as a smart location intelligence tool combining satellite imagery, spatial data, and AI models.