Inspiration

Every year, millions of tons of recyclable waste end up in landfills because people are unsure how to sort their trash. We wanted to build a tool that combines AI and sustainability to make recycling smarter, accessible, and fun for everyone.

What it does

WasteSort is an AI-powered app that classifies waste into categories like plastic, paper, glass, cardboard, metal, or general trash. Upload a photo of waste → AI instantly predicts the type. If recyclable → it finds nearby recycling centers using OpenStreetMap. If no centers are found → it gives clear recycling guidelines for that material.

How we built it

We trained a TensorFlow MobileNetV2 model on the TrashNet dataset for image classification. We built the interface in Streamlit, integrating both backend and frontend for quick deployment. We used OpenStreetMap’s Overpass API to fetch recycling center locations and displayed them with Folium maps. Finally, we deployed everything in a single interactive web app.

Challenges we ran into

Handling API limits — Google Maps required billing, so we switched to OpenStreetMap.

Accomplishments that we're proud of

Building a full-stack AI app in a single day. Creating a seamless workflow: upload → classify → get real-world recycling options. Designing a polished, demo-ready interface with real sustainability impact. Solving API and technical roadblocks quickly under time pressure.

What's next for WasteSort

Improving model accuracy with larger and more diverse datasets. Adding real-time camera support so users can classify trash directly. Making a User reward system which rewards users who actually go through with the recycling part.

Built With

• folium
• mobilenetv2
• overpass-openstreetmap
• python
• streamlit
• tensorflow
• trashnet