Securing potato Forms: A holistic approach to Classification of Early Blight and Late Blight Disease Management

Abstract

Potato cultivation faces significant economic challenges due to prevalent disease, particularly early blight and late blight. Early blight, caused by Alternaria solani, affects leaves and stems, potentially leading to defoliation and increased tuber infection risk. Late blight, attribute to phytophthora infestans, is a critical potato disease, capable of causing rapid crop failures. Timely disease detection is crucial for farmers to implement tailored treatments and minimizing financial losses. Early blight, associated witg fungi, and late blight, linked to specific microorganisms, require distinct control strategies, highlighting the need for accurate disease identification. This study introduces a deep convolutional neural network architecture with 14 layers, incorporating two primary convolutional layers for feature extraction with varying windows size, followed by two fully connected layers for classification. Augmentation techniques were applied to a dataset comprising 2152 images, resulting in a substantial enhancement in overall testing accuracy. The proposed architecture achieved an impressive mean testing accuracy of 98.43%. To ensure the robustness of the result, over six performance metrics were employed. This innovative deep learning approach holds promise for revolutionizing potato disease classification and advancing precision agriculture practices. The architecture implemented in a Jupyter notebook with GPU acceleration, demonstrated its effectiveness through meticulous testing and validation. The study underscores the potential of deep learning in providing accurate and efficient solutions for potato disease management, offering a transformative impact on agricultural practices and crop yield optimization.