Hybrid convolutional neural network based classification of bacterial, viral, and fungal diseases on tomato leaf images

Summary Tomatoes can get many diseases at every stage of the cultivation process depending on environmental and climatic factors. Tomato growers sometimes struggle to monitor leaves in order to detect these diseases. As a result, deep learning‐based systems for detecting diseases have been developed...

Full description

Saved in:
Bibliographic Details
Published in:Concurrency and computation 2022-02, Vol.34 (4), p.n/a
Main Authors: Cengil, Emine, Çınar, Ahmet
Format: Article
Language:English
Subjects:
Algorithms
Artificial neural networks
Classification
CNN
Datasets
Deep learning
Feature extraction
Fungal diseases
Image classification
Machine learning
machine learning algorithms
Medical imaging
mRMR
Neural networks
Object recognition
Optimization
Plant diseases
Redundancy
tomato leaf disease classification
Tomatoes
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Summary Tomatoes can get many diseases at every stage of the cultivation process depending on environmental and climatic factors. Tomato growers sometimes struggle to monitor leaves in order to detect these diseases. As a result, deep learning‐based systems for detecting diseases have been developed. It has demonstrated its worth by solving problems in a variety of fields, including classification. Convolutional neural network (CNN) is particularly effective in image classification, recognition, and detection. In this article, a hybrid‐based CNN model for the classification of diseases on tomato leaf images is proposed. The performance of the method has been examined by the tomato leaf disease detection dataset and Taiwan datasets. For the extraction of features, well‐known CNN architectures such as AlexNet, ResNet50, and VGG16 are used at first. The feature transfer method extracts features from the last fully connected layers of the architectures. After that, the minimum redundancy maximum relevance feature selection algorithm is applied to these features for optimization. The features gathered are concatenated. Concatenating features are classified by popular machine learning classification algorithms. With the proposed method, the highest performance values for the tomato leaf disease detection and Taiwan dataset show an accuracy of 98.3% and 96.3%, respectively.
ISSN:1532-0626
1532-0634
DOI:10.1002/cpe.6617