Smart & Efficient Waste Collection System

Inspiration

We wanted to create a sustainable project for a daily problem at our locality.

What it does

Monitors smart bins in real time, showing fill levels on a map. Classifies waste images uploaded by users into organic, recyclable, or e-waste using AI. Generates the most efficient garbage truck route to collect only those bins that are more than 80% full.

How we built it

Frontend: Built using Streamlit, integrated with Folium for maps and Plotly for charts. simulate_bins.py: Simulates real-time sensor data from bins, including location, fill level, and waste type. route_optimizer.py: Implements a greedy nearest-neighbor algorithm to optimize pickup routes for bins over 80% full. waste_classifier.py: Loads a machine learning model trained on labeled images to predict the type of waste in a photo. test_backend.py: Standalone backend tester to verify data generation and routing logic.

Challenges we ran into

No access to real IoT data: We had to simulate bin data manually, making sure it still represented real-world behavior. Model training: We didn’t have a ready-made labeled dataset for waste classification, so training a lightweight, accurate model was tricky. Route optimization logic: Balancing between performance and simplicity without external APIs like Google Maps was a constraint.

Accomplishments that we're proud of

Built a fully functional, interactive dashboard without needing real hardware. Integrated AI-based classification and route optimization in one unified app. Achieved real-time bin visualization with dynamic maps and charts.

What we learned

How to use Streamlit to build real-time, interactive dashboards. How to simulate realistic smart bin data without sensors. How to optimize routes programmatically using geospatial logic. How to train and use machine learning models in real-time apps

What's next for Smart & Efficient Waste Collection System

Deploy to cloud (e.g., Streamlit Community Cloud or AWS) so it can be used in real environments. Connect to real smart bins via IoT devices for live sensor data. Improve the classifier using deep learning and a larger image dataset. Integrate with Google Maps API to calculate real-world travel distances instead of Euclidean.

Built With

• follium
• pillow
• python
• streamlit