Plant Disease Diagnosis and Solution System Based on Neural Networks

Abstract

Plant diseases are one of the major factors affecting crop yield. Early identification of these diseases can improve productivity and save money and time for the farmer. This paper presents a novel technique to diagnose plant diseases using a mobile application. A Convolutional Neural Network (CNN) model was built and trained using MobileNetV2 architecture with the help of image processing techniques and transfer learning. A dataset comprising 87,000 images that contain 38 classes of diseases belonging to 14 different crops was used to train the model. The model achieved an accuracy of 98.69% and a loss of 0.5373. A mobile application was built in Android Studio with the help of a trained model. The mobile application built works without a need for a remote server. The application can identify the disease, gives information regarding the identified disease and also suggests necessary remedies to tackle the disease.