Inspiration
We noticed walking around campus that often recyclable items are not being recycled like they should and are being thrown into the trash. We wanted to build a system that uses AI to recognize and separate cans and plastic bottles to make recycling easier and more accurate for everyone.
What it does
SmartSort uses a camera and a trained AI model to detect whether an item is a plastic bottle or an aluminum can (recyclable) or waste basket. Once identified, the system activates a servo to either open the trash basket or the recycling basket.
How we built it
We built it by collecting our own datasets from different available datasets online and then training a yolov8 model on the mason hopper cluster, then we deployed our model on the raspberry pi. Using servo motors, we built a mechanism that directs items into the correct basket. The model runs on a small computer with a camera, and based on the prediction, the servos move accordingly to sort the item into the right place.
Challenges we ran into
One of the biggest challenges was getting the Raspberry Pi and Arduino to communicate effectively. We needed the Pi to send category detection results (bottle, can, or trash) to the Arduino, but since the Arduino primarily reads digital signals as 0 or 1, it was tricky to transmit multiple distinct categories reliably. Another challenge was designing our trash input system — we originally envisioned a circular, rotating platform using a stepper motor, but it didn't provide enough torque to handle the weight and motion of the system. This forced us to rethink and simplify our mechanical design.
Accomplishments that we're proud of
We’re proud of successfully building and training a model that can accurately detect and classify recyclable items. It was exciting to see the system recognize an object and trigger real-world actions. We integrated an LED indicator that lights up when a recyclable item is detected, followed by a servo arm that opens the recycling bin. While we originally planned to have three separate bins — one for bottles, one for cans, and one for general waste — we managed to get one functional bin working in time. Even though things didn’t go exactly as planned, pushing through the obstacles and staying determined was what really mattered to us.
What we learned
We learned that persistence is key — things won’t always go as planned, and that’s okay. What matters is staying flexible, adapting to the situation, and keeping a positive mindset. Sometimes, shifting gears is necessary to keep the project moving forward, and that's part of the creative process.
What's next for SmartSort: AI-Powered Recycling Bot
Next, we plan to integrate additional servo motors to control multiple trash bins , one for bottles, one for cans, and one for general waste — just as we originally envisioned. We also want to refine the design to make it more polished and visually appealing, fine-tune the software for even better accuracy, and eventually pitch the system for use in public spaces to promote smarter, cleaner recycling.
Built With python raspberry roboflow yolo
Log in or sign up for Devpost to join the conversation.