Development of an Automated Waste Sorter

Teoh, Wyne Jiet (2026) Development of an Automated Waste Sorter. Final Year Project (Diploma), Tunku Abdul Rahman University of Management and Technology.

Text Teoh Wyne Jiet_Thesis.pdf Restricted to Registered users only Download (10MB)

Abstract

Inefficient waste management has become a critical global challenge as it brings negative impact to the environment and human health. The traditional waste sorting method relies on manpower to sort manually and usually the sorting method is slow and error prone. Thus, this project aims to develop an automated waste sorter that sorts the household waste to increase the efficiency of waste sorting. This project proposes an automated waste sorter that uses machine vision to do waste classification and uses pan and tilt mechanism to sort the waste into the bin. This system uses Raspberry Pi 5 as the control system to perform waste classification and control the sorting mechanism. A waste classification model was built using TensorFlow Keras to train the waste classification model to identify waste such as plastic, paper, metal and other that is captured from the camera. A pan and tilt mechanism was used as the sorting mechanism to separate the waste into their corresponding bins. Garbage bin level detection is also incorporated into this system to prevent the waste from overflowing. This system aims to sort the waste accurately and increase the efficiency of sorting waste. This system offers a high accuracy EfficientNetV2B0 model that achieved a real-time accuracy of 93.33% with average latency of 38.25ms which also suitable for Raspberry Pi 5 as it has less model parameter