Advance Step to Protect Agriculture: A Deep Dive into Enhancing Food Security Through EfficientNet-B0 and XAI in Plant Disease Classification
Year: 2024
Type: conference-paper
Abstract: The global economy heavily relies on the agricultural industry, making crop health crucial for both survival and economic stability.Significant production losses are caused by plant diseases in agriculture.It is often difficult for experienced farmers to detect diseases with traditional methods, which are often insufficient.Modern frameworks that combine deep learning approaches and image segmentation look to be viable alternatives to prior techniques.The study begins with image augmentation and segmentation.The image segmentation is carried out using DeePLabV3+ for image preprocessing, followed by feature extraction.Subsequently, the classification ability of several deep learning models is studied, including Convolutional Neural Network (CNN), VGG16, MobileNetV2, AlexNet, InceptionV3, DenseNet201, and EfficientNet-B0, for the classification of 38 plant disease classes.Among the top-performing models, EfficientNet-B0 achieved 99.86% accuracy, 99.01%precision, 98.81% recall, and 99.34% F1 score.EfficientNet-B0 is also explained through SHAP (SHapley Additive exPlanations), which provides insight into how the model classifies leaf diseases.Based on the results of the implementation of EfficientNet-B0, it is possible to significantly reduce crop destruction caused by plant diseases, which improves food security and economic stability.
Authors: Sadia Nawar, Altaf Uddin, Arnob Saha, Md. Hamid Hosen, Safa Asgar, Linkon Chowdhury
Institution: East Delta University
Language: English
Topic: Smart Agriculture and AI
Subfield: Plant Science
Domain: Life Sciences
SDG: Zero hunger
Open Access status: hybrid
OpenAlex ID: W4411098863