Investigation and comparison of different CNN architectures on tomato leaf disease prediction using deep learning

Abstract

The tomato crop is the most common vegetable plant in the Indian market which is economically valuable and cultivated in enormous volumes. The productivity diminishes to large scale due to diseases caused by pests, and pathogens. The tomato diseases are of different forms that affect plant's root, stem, leaves and so on. Monitoring the tomato crops for detection of diseases plays a crucial role in optimum production. The existing method for detecting tomato leaf diseases is simply the visual observation by agricultural experts and plant pathologists. It is time consuming and also costs high. Sometimes, they often fail to diagnose specific diseases leading to inaccurate assumptions. The proposed method uses Convolutional Neural Network to identify the diseases of tomato leaves with the help of features that are extracted from the images. It implements three different CNN architectures and experimented with three different Optimizers. One of the CNN architecture is LeNet, it achieves an accuracy of around 90-92%. Other architecture is AlexNet where it achieves an accuracy of around 92-97%.The third architecture is VGG16, it achieves an accuracy of around 90- 94%.