Tomato crop disease classification using pre-trained deep learning algorithm

Abstract

The wide scale prevalence of diseases in tomato crop affects the production quality and quantity. In order to counteract the problem early diagnosis of diseases using a fast reliable nondestructive method will benefit the farmers. In this study images of tomato leaves (6 diseases and a healthy class) obtained from PlantVillage dataset is provided as input to two deep learning based architectures namely AlexNet and VGG16 net. The role of number of images and significance of hyperparameters namely minibatch size, weight and bias learning rate in the classification accuracy and execution time have been analyzed.