ECONET

EcoNet: Turning Waste into Worth, Dignity into Pay.

Comment 1

Inspiration

Growing up in India, we saw how ragpickers sort mountains of waste every day, earning barely enough to survive while facing unsafe and undignified conditions. At the same time, cities struggle with landfill overflow and poor waste segregation. EcoNet was inspired by the idea that AI can not only make recycling smarter, but also give dignity and fair pay to those who hold our waste economy together.

What it does

EcoNet is a deep learning–powered web app that classifies waste images into categories such as paper, plastic, metal, and glass. Citizens can upload an image, and the model instantly predicts the category. This enables efficient waste segregation, promotes recycling, and lays the foundation for connecting ragpickers directly with organized recycling systems.

How we built it

We trained a convolutional neural network (CNN) using TensorFlow/Keras on waste classification datasets (Garbage Classification + TrashNet). Data preprocessing included resizing, normalization, and augmentation. The model achieved ~85.9% training accuracy and ~90.1% test accuracy. We then deployed the trained model as a web app using Gradio, making it simple for anyone to upload an image and see predictions in real time.

Challenges we ran into

Managing dataset structures and preprocessing due to messy folder hierarchies.

Colab session resets, which meant reconnecting and re-downloading datasets.

Streamlit/LocalTunnel setup issues for deployment, which we later solved by using Gradio for a stable demo.

Coordinating tasks across a team where some members were non-coders but contributed in design and presentation.

Accomplishments that we're proud of

Building a working AI model with 90%+ test accuracy within the hackathon timeline.

Successfully deploying the model into a simple, usable web app.

Turning an abstract idea into a tangible prototype that connects AI to a real-world sustainability problem.

Collaborating as a team where everyone contributed in different ways — coding, design, and storytelling.

What we learned

How to preprocess datasets and train CNNs effectively for image classification.

How deployment works with tools like Gradio and Streamlit.

The importance of backup plans when tools fail (switching from LocalTunnel to Gradio).

Most importantly, how AI can be linked to social impact and SDGs, not just tech demos.

What's next for ECONET

Expand classification to more waste categories, including organic waste and hazardous waste.

Integrate an impurity level detector to help recycling centers save time and costs.

Partner with NGOs and recycling industries to connect ragpickers directly with buyers.

Build a gamified reward system (EcoPoints) to encourage citizens to recycle through the platform.

Scale the model and platform for real-world testing beyond hackathons.

Built With

Share this project:

Updates