1. RSUIR at Rangsit University
2. RSU Digital Collection
3. วิทยาลัยวิศวกรรมศาสตร์ COLLEGE OF ENGINEERING
4. Eng-ECE-M-Thesis

Title:	Mango maturity classification using machine learning
Authors:	Xiangchen Dong
metadata.dc.contributor.advisor:	Rong Phoophuangpairoj
Keywords:	Machine learning;Mango -- Temperature;Deep learning (Machine learning) -- Industrial applications
Issue Date:	2024
Publisher:	Rangsit University. Library
metadata.dc.description.other-abstract:	Mangoes, predominantly cultivated in tropical and subtropical regions, are cherished for their sweet and sour taste, pleasant aroma, and rich vitamin content. This study focused on classifying mango ripeness using various machine learning classifiers. Images of mangoes at different ripeness stages were collected and used to train classifiers, including Convolutional Neural Network (CNN), MobileNet, ResNet50, and VGG16. The experimental results demonstrated that CNN, MobileNet, ResNet50, and VGG16 achieved accuracies of 81.30%, 85.11%, 73.66%, and 90.08%, respectively. VGG16 attained the highest classification accuracy, with classification accuracies from class 0 to class 5 being 98.85%, 98.85%, 95.80%, 95.80%, 95.42%, and 95.42%, respectively. Building on these results, the second part of the study utilized the top-performing VGG16 model to output softmax layer data from mango images. This softmax data, combined with the mangoes’ RGB and Lab color data, was used to make predictions via linear regression and artificial neural networks (ANNs). The comparative analysis revealed that ANN generally yielded better predictive performance
Description:	Thesis (M. Eng. (Electrical and Computer Engineering)) -- Rangsit University, 2024
metadata.dc.description.degree-name:	Master of Engineering
metadata.dc.description.degree-level:	Master's Degree
metadata.dc.contributor.degree-discipline:	Electrical and Computer Engineering
URI:	https://rsuir-library.rsu.ac.th/handle/123456789/3163
metadata.dc.type:	Thesis
Appears in Collections:	Eng-ECE-M-Thesis

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.