Comparative Analysis of Convolutional Neural Network Models for Solid Waste Categorization

• Author(s): Tamuno-omie J. Alalibo ; Nkolika O. Nwazor
• Paper ID: 1704821
• Page: 1396-1405
• Published Date: 21-07-2023
• Published In: Iconic Research And Engineering Journals
• Publisher: IRE Journals
• e-ISSN: 2456-8880
• Volume/Issue: Volume 6 Issue 12 June-2023

Abstract

This paper is on the comparative analysis of various convolutional neural network (CNN) models that could be used in the classification of solid waste into different categories for recycling. Effective solid waste classification is a crucial aspect of waste disposal and recycling processes to maintain a sustainable environment. In recent years, CNNs have emerged as a powerful tool for solid waste classification, thanks to their ability to learn features from images and classify objects accurately. In this paper, the MATLAB platform was used to train eight pre-trained CNN models for solid waste classification. The models are AlexNet, GoogleNet, ResNet18, MobileNetV2, ResNet50, ResNet101, EfficientNetB0 and InceptionV3. Each model was trained for 5 epochs, 7 epochs, and 10 epochs respectively. The TrashNet dataset was used. The dataset was split into 75% and 25% for the training set and validation set respectively. After training for 5 epochs, ResNet50 achieved validation accuracy of 90.95%, ResNet101 had 90.48%, while the rest achieved accuracies ranging from 64% to 89%. After training for 7 epochs, ResNet50 got an accuracy of 92.06%, ResNet101 and InceptionV3 both had an accuracy of 91.59%, while EfficientNetB0 improved to 90.63%. After 10 epochs, ResNet101 achieved an accuracy of 92.38%, ResNet50 got an accuracy of 92.02%, while EfficientNetB0 and InceptionV3 achieved an accuracy of 91.43%. From the results, ResNet-50 and ResNet-101 achieved higher validation accuracy. In real-world applications, these models can be used in smart bin waste collection, smart waste recycling, and waste management in general.

Keywords

Convolutional Neural Network, Waste Image Classification, TrashNet, ResNet50, Recycling

Citations

IRE Journals:
Tamuno-omie J. Alalibo , Nkolika O. Nwazor "Comparative Analysis of Convolutional Neural Network Models for Solid Waste Categorization" Iconic Research And Engineering Journals Volume 6 Issue 12 2023 Page 1396-1405

IEEE:
Tamuno-omie J. Alalibo , Nkolika O. Nwazor "Comparative Analysis of Convolutional Neural Network Models for Solid Waste Categorization" Iconic Research And Engineering Journals, 6(12)