Grape Leaf Diseases Detection Using Deep Learning

Abstract

Even though a lot of research has been done in this area, classifying grape leaf diseases remains a very difficult task in the agricultural industry. This task is made more challenging by the spread of leaf diseases and the variations in their structural makeup. The scale-invariant feature transform (SIFT) and the histogram of directed gradients are two examples of manually created features that are used as the basis for practically all classifications nowadays (HOG). Convolutional Neutral Network (CNN), a feature learning technique, will be used in this instance. Here, the major objective is to train a model that has already been trained to find and extract the features that are essential for the given task. It finds several grape leaf diseases such as Black Rot, Esca, Leaf Blight, etc. Our goal is to create a deep learning- and image-based system for detecting grape leaf disease. The Convolutional Neural Network (CNN) method and Transfer Learning Implementation Technique are used in this approach to learning from a large number of unlabeled picture patches and a large number of healthy and infected leaf images. To detect the presence of infection in the grape leaf during the study, we will use some previously photographed diseased leaves and analyze the data using the deep learning algorithm Convolutional Neutral Network (CNN). A deep learning method will be applied to a sample dataset of the infected leaf in this project (CNN). For model building, we'll employ a few Tensor Flow preprocessed models. Because this system will be able to readily identify various diseases, we hope that we will be able to create a system that will offer us more accurate results for detecting grape leaf diseases. Our approach aims to decrease grape leaf illnesses before they become contagious infected.