Automated Waste Image Classification with Weighted Scoring Using MobileNetV2 on the OLSAM Platform

Abstrak

This study presents the development of an automated waste image classification system for the OLSAM platform to enhance community participation in waste management. The objective is to integrate a lightweight CNN-based classifier with a weighted point calculation mechanism for five waste categories. A dataset of 1,500 images was used, split into 80% training, 10% validation, and 10% testing. The MobileNetV2 architecture was applied to perform image classification, while a weighted reward mechanism assigned points based on the detected waste type and its weight. The model achieved its best performance at epoch 65, reaching an accuracy of 96.67% and a weighted F1-score of 0.97. These results indicate that combining CNN-based recognition with a weighted point system effectively supports user engagement and promotes sustainable waste-sorting behavior within community waste management systems.