Grape Leaf Disease Diagnosis System Using Fused Deep LearningFeatures Based System

Abstract

Rapid and accurate grape leaf disease detection be critical into grape productivity as well as maybe even caliber. The study proposes a quick and precise detection technique for grape leaf diseases according to fused deep features collected from both Convolutional neural networks and support vector machines.Three different forms of cutting-edge seven distinct Deep feature layer types, CNN networks, three distinct deep feature fusion techniques, and a multi-class over that,or explored in the study, which was based on an open dataset. Photos were first scaled to fit the CNN network's input specifications, and then, the input pictures' deep characteristics were retrieved using the CNN network's unique deep feature layer. The effective classification feature information was then increased by fusing two distinct types of deep features from separate networks. Finally, the merged deep features were utilized to train a multi-class SVM classifier. The experimental findings on the publicly available dataset demonstrate that fusing deep features in any way may improve classification performance over utilizing a single kind of deep feature. In comparison to the other two fusion approaches, the best classification outcomes may be obtained by directly concatenating the Fc1000 deep feature derived from ResNet50 and ResNet101, with an F1 score of 99.82%. Additionally, SVM classifier developed with the suggested technique is capable of achieving classification performance equivalent to Compared to having trained a CNN model for tens of minutes,utilizing the CNN model straight up requires less training time of less one second. In the experimental findings show that the approach suggested in this research may identify grape leaf diseases quickly and accurately while also meeting the demands of real agricultural operations.