An effective approach for plant leaf diseases classification based on a novel DeepPlantNet deep learning model

Area of Science:

• Agricultural Science
• Computer Science
• Plant Pathology

Background:

• Plant diseases pose a significant threat to global food security and agricultural productivity.
• Current methods for plant disease diagnosis are often time-consuming, labor-intensive, and lack accuracy.
• The need for automated, efficient, and reliable plant disease detection systems is critical.

Purpose of the Study:

• To develop and evaluate an efficient, novel, and lightweight deep learning (DL) architecture for plant leaf disease prediction and classification.
• To address the limitations of traditional plant disease diagnosis methods through an automated approach.

Main Methods:

• A novel deep learning architecture named DeepPlantNet was proposed, featuring 28 learned layers (25 convolutional and 3 fully connected).
• The DeepPlantNet model incorporates Leaky ReLU activation, batch normalization, fire modules, and a combination of 3x3 and 1x1 filters.
• The framework is designed for efficient categorization of plant disease images into multiple classes.

Main Results:

• The DeepPlantNet model achieved high classification accuracy, reaching 98.49% for an eight-class scheme and 99.85% for a three-class scheme.
• Experimental results demonstrated the superiority of DeepPlantNet compared to alternative methods.
• The model effectively categorizes plant diseases into ten distinct groups, including Apple Black rot, Cherry Powdery mildew, and Maize common rust.

Conclusions:

• DeepPlantNet offers a highly accurate and efficient solution for automated plant leaf disease identification.
• The proposed DL model can significantly reduce financial losses and improve agricultural output by enabling early disease detection.
• This technology empowers farmers and agricultural professionals with a tool for timely and effective disease management.