Open Access

ARTICLE

Optimization of Deep Learning Model for Plant Disease Detection Using Particle Swarm Optimizer

by Ahmed Elaraby^1,*, Walid Hamdy², Madallah Alruwaili³

1 Department of Computer Science, Faculty of Computers and Information, South Valley University, Qena, Egypt
2 Department of Math., & Computer Science Faculty of Science, Port Said University, Egypt
3 College of Computer and Information Sciences, Jouf University, Sakaka, Aljouf, Kingdom of Saudi Arabia

* Corresponding Author: Ahmed Elaraby. Email:

(This article belongs to the Special Issue: Recent Advances in Deep Learning and Saliency Methods for Agriculture)

Computers, Materials & Continua 2022, 71(2), 4019-4031. https://doi.org/10.32604/cmc.2022.022161

Received 29 July 2021; Accepted 20 October 2021; Issue published 07 December 2021

Abstract

Plant diseases are a major impendence to food security, and due to a lack of key infrastructure in many regions of the world, quick identification is still challenging. Harvest losses owing to illnesses are a severe problem for both large farming structures and rural communities, motivating our mission. Because of the large range of diseases, identifying and classifying diseases with human eyes is not only time-consuming and labor intensive, but also prone to being mistaken with a high error rate. Deep learning-enabled breakthroughs in computer vision have cleared the road for smartphone-assisted plant disease and diagnosis. The proposed work describes a deep learning approach for detection plant disease. Therefore, we proposed a deep learning model strategy for detecting plant disease and classification of plant leaf diseases. In our research, we focused on detecting plant diseases in five crops divided into 25 different types of classes (wheat, cotton, grape, corn, and cucumbers). In this task, we used a public image database of healthy and diseased plant leaves acquired under realistic conditions. For our work, a deep convolutional neural model AlexNet and Particle Swarm optimization was trained for this task we found that the metrics (accuracy, specificity, Sensitivity, precision, and F-score) of the tested deep learning networks achieves an accuracy of 98.83%, specificity of 98.56%, Sensitivity of 98.78%, precision of 98.67%, and F-score of 98.47%, demonstrating the feasibility of this approach.

---

Keywords

---

Citations