Deep Learning-Based Disease Identification and Classification in Potato Leaves

Abstract

This research endeavored to deploy deep learning models for accurate illness detection within potato crops, a crop of global economic significance. Four specific deep learning models, namely VGG16, EfficientNet B4, InceptionV3, and Inception ResNetV2 were trained using a comprehensive dataset constituted of images of both healthy and diseased potatoes. The performance of these models far surpassed traditional methods of visual inspection. Among the models evaluated, the EfficientNet B4 model demonstrated the highest level of accuracy, achieving a perfect score of 100%. VGG16 followed with an accuracy of 99%, Inception V3 at 98%, and Inception ResNet V2 at 94%. The results yield significant potential for enhanced crop management methodologies, facilitating a considerable reduction in economic losses linked to potato diseases. The study's findings corroborate the transformative capacity of artificial intelligence, particularly deep learning models, for the innovation of agricultural practices. This is pertinent for upholding food security in light of the increasing challenges posed by plant diseases.