An Improved Convolutional Neural Network (CNN) for Disease Detection and Diagnosis for Multi-crop Plants

Agriculture is one of the largest sectors that contribute to the economic growth of countries, including Malaysia. However, plant diseases affect the quality of the harvest and impede farmers’ maximum yield output. Therefore, early detection of diseases in plants is vital to curb infection, reduce f...

Full description

Saved in:
Bibliographic Details
Main Authors: Florence Choong Chiao Mei, Bryan Ng Jan Hong
Format: Article
Language:English
Published: MMU Press 2025-03-01
Series:Journal of Engineering Technology and Applied Physics
Subjects:
Online Access:https://journals.mmupress.com/index.php/jetap/article/view/1158/766
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849736509054255104
author Florence Choong Chiao Mei
Bryan Ng Jan Hong
author_facet Florence Choong Chiao Mei
Bryan Ng Jan Hong
author_sort Florence Choong Chiao Mei
collection DOAJ
description Agriculture is one of the largest sectors that contribute to the economic growth of countries, including Malaysia. However, plant diseases affect the quality of the harvest and impede farmers’ maximum yield output. Therefore, early detection of diseases in plants is vital to curb infection, reduce food waste, and reduce their carbon footprint. However, many detection methods are complex, require high computational power and time to perform the required analysis and focus only on a particular species or strain of the disease. These requirements would likely deter most users in remote areas or poorer economic states. This paper proposes a convolutional neural network to determine multi-class plant diseases that is memory efficient, has a small trainable parameter number, and is compact enough to work even on mobile devices. The plant images were pre-processed to ensure that they were validated accurately and to minimise overfitting. Then, the proposed convolutional neural network was trained using a publicly available dataset consisting of 54306 images, followed by validation and testing. Finally, the completed model is saved, and the data obtained is transferred to a cloud network using wireless sensor networks. The proposed method obtained 96.87% accuracy with 100 epoch training iterations, rivalling famous architectures such as VGG16 and MobileNetV2. The experimental results demonstrate the feasibility and robustness of the method for disease detection in multi-crop plants.
format Article
id doaj-art-003cca7555814bd597fe2faff38bb640
institution DOAJ
issn 2682-8383
language English
publishDate 2025-03-01
publisher MMU Press
record_format Article
series Journal of Engineering Technology and Applied Physics
spelling doaj-art-003cca7555814bd597fe2faff38bb6402025-08-20T03:07:14ZengMMU PressJournal of Engineering Technology and Applied Physics2682-83832025-03-017171410.33093/jetap.2025.7.1.2An Improved Convolutional Neural Network (CNN) for Disease Detection and Diagnosis for Multi-crop PlantsFlorence Choong Chiao Mei0https://orcid.org/0000-0002-6958-8725Bryan Ng Jan Hong1Engineering andPhysical Sciences, Heriot-Watt University Malaysia, Putrajaya, Malaysia.Engineering andPhysical Sciences, Heriot-Watt University Malaysia, Putrajaya, Malaysia.Agriculture is one of the largest sectors that contribute to the economic growth of countries, including Malaysia. However, plant diseases affect the quality of the harvest and impede farmers’ maximum yield output. Therefore, early detection of diseases in plants is vital to curb infection, reduce food waste, and reduce their carbon footprint. However, many detection methods are complex, require high computational power and time to perform the required analysis and focus only on a particular species or strain of the disease. These requirements would likely deter most users in remote areas or poorer economic states. This paper proposes a convolutional neural network to determine multi-class plant diseases that is memory efficient, has a small trainable parameter number, and is compact enough to work even on mobile devices. The plant images were pre-processed to ensure that they were validated accurately and to minimise overfitting. Then, the proposed convolutional neural network was trained using a publicly available dataset consisting of 54306 images, followed by validation and testing. Finally, the completed model is saved, and the data obtained is transferred to a cloud network using wireless sensor networks. The proposed method obtained 96.87% accuracy with 100 epoch training iterations, rivalling famous architectures such as VGG16 and MobileNetV2. The experimental results demonstrate the feasibility and robustness of the method for disease detection in multi-crop plants.https://journals.mmupress.com/index.php/jetap/article/view/1158/766plant diseaseagricultureconvolutional neural networkimage processingwireless sensor network
spellingShingle Florence Choong Chiao Mei
Bryan Ng Jan Hong
An Improved Convolutional Neural Network (CNN) for Disease Detection and Diagnosis for Multi-crop Plants
Journal of Engineering Technology and Applied Physics
plant disease
agriculture
convolutional neural network
image processing
wireless sensor network
title An Improved Convolutional Neural Network (CNN) for Disease Detection and Diagnosis for Multi-crop Plants
title_full An Improved Convolutional Neural Network (CNN) for Disease Detection and Diagnosis for Multi-crop Plants
title_fullStr An Improved Convolutional Neural Network (CNN) for Disease Detection and Diagnosis for Multi-crop Plants
title_full_unstemmed An Improved Convolutional Neural Network (CNN) for Disease Detection and Diagnosis for Multi-crop Plants
title_short An Improved Convolutional Neural Network (CNN) for Disease Detection and Diagnosis for Multi-crop Plants
title_sort improved convolutional neural network cnn for disease detection and diagnosis for multi crop plants
topic plant disease
agriculture
convolutional neural network
image processing
wireless sensor network
url https://journals.mmupress.com/index.php/jetap/article/view/1158/766
work_keys_str_mv AT florencechoongchiaomei animprovedconvolutionalneuralnetworkcnnfordiseasedetectionanddiagnosisformulticropplants
AT bryanngjanhong animprovedconvolutionalneuralnetworkcnnfordiseasedetectionanddiagnosisformulticropplants
AT florencechoongchiaomei improvedconvolutionalneuralnetworkcnnfordiseasedetectionanddiagnosisformulticropplants
AT bryanngjanhong improvedconvolutionalneuralnetworkcnnfordiseasedetectionanddiagnosisformulticropplants