-
Website (Frontend)
Inspiration
We wanted to work on a problem that actually affects our community. Recycling seems simple, but a lot of recyclable material gets contaminated by things that are put in the wrong bin. We thought AI could be used to help with this problem instead of just being another app or chatbot.
What it does
EcoSort uses a camera and an AI object detection model to identify different types of waste and classify them as recyclable, compost, or garbage. The goal is to eventually use this system in recycling facilities to help catch incorrectly sorted items before they contaminate the recycling stream.
How we built it
We started by using a webcam connected to a laptop and trained/tested an object detection model using TensorFlow. The camera looks at an item and the model predicts what type of waste it is.
We also planned the system around an Arduino Uno R3 so the AI could eventually control a physical sorting mechanism. This would allow the project to go beyond just recognizing an object and actually sort it.
Challenges we ran into
One of our biggest challenges was getting the AI to correctly recognize different objects. Waste can look very different depending on the angle, lighting, and condition of the item.
Another challenge was figuring out how to connect the software side of the project to actual hardware. We had to think about how the camera, AI model, Arduino, and sorting mechanism would all work together.
Accomplishments that we're proud of
We're proud that we were able to turn our idea into a working prototype instead of just having a concept. We got the AI to recognize waste using a webcam and started working toward connecting it to physical hardware.
We're also proud that our project focuses on a real problem and could potentially be scaled up for use in larger recycling facilities.
What we learned
We learned a lot about computer vision, object detection, TensorFlow, and working with hardware. We also learned that building something that works in real life is a lot harder than just having the idea. Getting the software and hardware to work together took a lot of testing and troubleshooting.
What's next for EcoSort
Our next step is to finish the physical sorting system using the Arduino and improve the accuracy of the AI model. Eventually, we want EcoSort to be designed for Material Recovery Facilities (MRFs), where it could help identify contamination and improve the amount of material that actually gets recovered and recycled.
