Waste Classification Model Optimization with Modified MobileNetV3 for Efficient Waste Management

Abstract

The increase in population and economic activity has a significant impact on the amount of waste. Data in 2023 states that waste in Indonesia still cannot be managed properly. One solution to overcome this problem is through recycling with waste sorting as a crucial stage. This research develops a waste classification model using modified MobileNetV3S. The classification process is performed using Convolutional Neural Net- work (CNN) method and parameter fine-tuning. This model is able to classify five different categories of waste, namely plastic bottles, leaves, plastic sheets, paper, and metal. These categories were chosen by considering the common practice in waste classification for recycling purposes. The results show that the validation accuracy
reaches 96.2% with a loss value of 0.049. These results can significantly contribute to better and sustainable waste management efforts.