Automated Plant Disease Detection Using Efficient Deep Ensemble Learning Model for Smart Agriculture

Automated Plant Disease Detection Using Efficient Deep Ensemble Learning Model for Smart Agriculture

R. Karthick Manoj, Aasha Nandhini S., T. Sasilatha
Copyright: © 2024 |Pages: 19
DOI: 10.4018/979-8-3693-0639-0.ch014
OnDemand:
(Individual Chapters)
Available
$37.50
Current Special Offers
No Current Special Offers
TOTAL SAVINGS: $37.50

Abstract

Early diagnosis of plant diseases is essential for successful plant disease prevention and control, as well as agricultural production management and decision-making. In this research, an efficient weighted average deep ensemble learning (EWADEL) model is used to detect plant diseases automatically. Transfer learning (TL) is a technique used to enhance existing algorithms. The performances of several pre-trained neural networks with DL such as ResNet152 DenseNet201, and InceptionV3, in addition to the usefulness of a weighted average ensemble models, are demonstrated for disease linked with leaf identification. To that aim, a EWADEL methodology is being researched in order to construct a robust network capable of predicting 12 different diseases of apple, Pomegranate, and tomato crops. Several convolutional neural network architectures were examined and ensemble to increase predictive performance using the EWADEL. In addition, the proposed approach included an examination of several deep learning models and developed EWADEL models.
Chapter Preview
Top

Literature Survey

This paper Javidan et.al (2023) provides a unique weighted majority voting ensemble strategy for detecting tomato leaves by distinguishing red, green, and blue pictures. As basic classifiers, six machine learning approaches were used. The proposed methods were then employed to improve sickness classification, with precision rates of 93.49% and 95.58%, correspondingly. The suggested technique's performance was compared to two well-known DL algorithms, which produced poor results. The suggested framework based on weighted majority voting beat the underlying ML, according to the study's findings. An infection identification method for crops has been developed employing either diseased and healthy leaves from different plant classes during this investigation Kondaveeti, et.al (2023). The fundamental models achieved accuracy levels below 90% based the crop illness data. The core algorithms now contain hard and soft voting categories to increase the system's accuracy. Soft voting entails examining the anticipated accuracy of each foundation methodology and selecting the way with the highest average weight as the final forecast.

Complete Chapter List

Search this Book:
Reset