Overview

Illustration of motivation figuring new idea

This project presents an automated disposable cup separation and return system designed to address Korea's 2022 disposable cup deposit policy. The system automatically classifies cup materials (paper, plastic, cup holders) using deep learning, removes lids, washes cups, and stacks them by material type. Using EfficientNetB0 transfer learning and real-time OpenCV processing, the system achieves over 99% classification accuracy at 8+ FPS, reducing labor burden on café owners while promoting proper recycling.

Background and goals

In June 2022, Korea's Ministry of Environment implemented a deposit system for café disposable cups to combat pollution. This policy created significant labor and workforce challenges for café owners who must handle cup returns, separation, and washing.

Goals:

• Automate cup material recognition (paper, plastic, cup holders) using image classification
• Automatically remove lids and wash cups
• Stack cups by material type and transfer to storage bins when full
• Minimize additional manpower needed for cup return processing
• Support users in proper waste separation with minimal effort

System Architecture

Overview

The system consists of:

• MCU: STM32F446 board for sensor and motor control
• Deep learning: Transfer learning to identify disposable cup materials
• Real-time processing: OpenCV for frame-by-frame material prediction
• Communication: Serial communication to STM32F446, socket communication to UI