WHEN AI SEES LITTER

Foreign residents and visitors in Japan often feel confused about trash disposal because each city has its own complex rules. Even small mistakes, such as putting a lithium battery in burnable trash, can lead to neighbor disputes or serious hazards at recycling plants. We built an AI-powered trash classification app to address this problem. It detects items in real time through a smartphone camera and then shows clear instructions based on the city’s rules. Our goal is safer waste disposal, fewer conflicts, and truly sustainable cities where materials are recycled properly.

The app looks at the item in your hand and tells you exactly how to dispose of it for your city. We trained YOLOv11 object detection models on the TACO trash dataset and improved performance with cropped and augmented images. The model recognizes items such as PET bottles, cans, and styrofoam. We map these to city-specific categories, for example Shibuya burnable, non burnable, and recyclable. The model is integrated into a Streamlit app with a simple interface that provides real-time classification and disposal instructions. The app also links to updated SDG articles to build everyday sustainability awareness.

Next we will add more municipalities, increase accuracy with localized data, support multiple languages, and package a mobile version that works offline. The vision is a friendly guide that removes confusion and turns correct disposal into a simple daily habit for everyone.