W.A.S.T.E. · Mateusz Hibner

Animated classification walkthrough — Quick GIF preview of the classification and lid-opening flow.

W.A.S.T.E. prototype mockup — Prototype shell showing dual compartments and signage.

Overview

W.A.S.T.E. is a CNN-powered trash can that recognizes common waste streams and automatically opens the matching compartment. By guiding people with visual cues and hands-free lids, it lowers the cognitive load of sorting trash while keeping the interaction playful and intuitive.

How it works

The system runs on a Raspberry Pi 5 with a Camera Module 2 and an HC-SR04 ultrasonic sensor that wakes the camera when someone approaches. A fine-tuned YOLO 11s classifier (trained on 13k images) evaluates short bursts of frames; if enough frames cross a 60% confidence threshold, the corresponding SG90 servo opens either the paper lid or the plastic/metal lid. LEDs provide immediate feedback during inference or rejection.

Desktop demo showcasing the fine-tuned model running end-to-end.