What: An automated waste-sorting system that utilizes computer vision and machine learning to classify waste into landfill, recyclable, or compostable bins. Developed at Cal Hacks 11.0 in San Francisco.

Why: We were inspired when we saw the Green Mary team manually sorting incorrectly disposed trash at an event. After speaking with their employees, we learned about San Francisco’s strict Zero Waste policy, which mandates large events hire waste management companies to ensure compliance. Failure to meet this policy could result in event organizers being banned from hosting future events, prompting us to develop an automated solution to improve sorting accuracy and efficiency.

How: Using Fusion 360, I designed a custom 3D-printed funnel system that directs waste to the correct bin based on classifications from Ultralytics' YoloV11 object detection model running on a Raspberry Pi. The funnel's design was optimized for precise measurements to ensure smooth waste flow. Testing tolerances on a smaller scale before the final print was crucial to saving the limited time we had.

Results: We successfully developed a functional prototype within 36 hours and showcased our project to judges, mentors, and fellow hackers. This experience helped me refine my skills in tolerancing for 3D design, rapid prototyping, and team collaboration.

Links: DevPost, Demo Video