Inspiration

Did you know that the US is on track to run out of landfill space by 2036, and roughly 80% of items in landfills could actually be recycled? We wanted to take on this challenge while also diving into computer vision and AI, using new technologies to address new problems. RecycleOrNot is our way of making recycling easy and accessible, one scan at a time.

What it does

RecycleOrNot is a user-friendly web application that uses your camera to scan and classify items in real time, letting you know whether they are recyclable or not.

How we built it

We built RecycleOrNot using Flask for the backend and React for the user interface. For object detection and classification, we integrated OpenCV and Ultralytics YOLO. The Python Flask backend processes image data captured through the camera, classifies items as recyclable or non-recyclable, and sends results to the React front end for display.

Challenges we ran into

Initially we were trying to build our RecycleOrNot using a pre-trained machine learning model called "resnet50v2." Although we were almost successful and even finished training the data, our program would result in extremely inconsistent predictions. Hence, our group decided to scrap this program and utilize the computer vision hackpack provided as our starting point.

Accomplishments that we're proud of

Our team had no previous experience in terms of dealing with computer vision and artificial intelligence for object classification. Hence why we are extremely proud of being able to incorporate all of these different technologies together in a full-stack app.

What we learned

We learned how to incorporate computer vision and artificial intelligence together to be able to create an object classification model that's able to work with moving items in real time. Aside from that, most of our team had no experience in building a full stack app, we were able to learn how to use Flask to send and receive requests from React.

What's next for Recyclable or Not?

Currently, Recyclable or Not's model is not as fine tuned as we'd like it to be. In the future, we would aim for Recyclable or Not to be able to classify things more accurately.

Share this project:

Updates