An Accurate and Automated plant disease detection system using transfer learning based Inception V3Model

Rice is one of the most widely grown crops across the globe, and it is susceptible to a variety of illnesses at various phases of production. Crop pathogens have a devastating influence on food safety, as well as a considerable loss in both the quantity and quality of production. Plant infections can damage the crop entirely in severe circumstances. Farmers' poor understanding makes it extremely difficult for them to visually diagnose these diseases. As a result, in agriculture domain, automatic recognition and diagnosis of crop diseases are greatly needed. Several approaches for solving this problem have been offered, including deep learning emerging as the best choice because of its excellent results. Recent advancements in Deep Learning have demonstrated that Automatic image identification methods based on CNN models can be extremely useful in such situations. We examine transfer learning of pre-trained deep CNN architectures for the recognition of infections found in plant leaves. We have utilized a pre-trained network trained on huge datasets and then applying it to our task using specific dataset. In our technique, we used the Inception V3pre-trained deep CNN model. Input samples of 2550 infected leaf images were acquired from the rice leaf dataset consisting of 5 disease classes. Contrast stretching procedure was applied on collected image samples to get good visual characteristics of the images. Image augmentation was used to avoid overfitting of the model during the training. Result analysis illustrated that the proposed strategy produced promising result with a 100 percent overall accuracy.