Cycle AI: Using Artificial Intelligence to Classify Recyclables

Cycle AI: Using Artificial Intelligence to Classify Recyclables

Title: Cycle AI: Using Artificial Intelligence to Classify Recyclables
Group: Michael Fourie, James Gleave, Luis Mangubat, Jean Yaacoub
Course: This project was for QHacks and The Mayor’s Innovation Challenge

Description: Cycle AI is a mobile application that utilizes the power of artificial intelligence to classify recyclables, organics, hazardous waste, as well as regular waste allowing users to properly dispose of their waste. The original model built during QHacks utilized a standard convolutional neural network that could classify only one piece of waste in an image. After receiving some constructive feedback from the City of Kingston judges, the team created an entirely new more advanced convolutional neural network that has the ability to isolate and mask images. This means that users can now take a photo of as much waste as they would like, and Cycle would be able to classify each piece of waste in that photo. This entire project was written in Python, using Kivy for the front-end, and TensorFlow, Keras, and OpenCV for the back-end. Cycle has a built-in achievement system, granting users rewards such as using the app consistently, scanning every category of recyclables, and reaching 100 total scans. This is to incentivize users to use the app more often, furthering their knowledge of proper recycling habits.

Watch the group’s pitch for the Mayor’s Innovation Challenge (starts at 9:50).

More information can be found on the Cycle AI Devpost page.



Doug Martin

Systems Specialist, School of Computing, Queen's University