Tomato diseases Classification Based on VGG and Transfer Learning

Abstract

Information technologies can introduce important innovation in human life and daily activities. Among the most important innovations developed in recent years, those concerning the agriculture are particularly relevant even from an economic point of view.The main advantage is the cross-analysis of environmental, climatic, and cultural factors, which allows establishing the irrigation and nutritional needs of crops, preventing pathologies, identifying weeds before they proliferate.Specifically, the main contribution of this work consists in the use of three convolutional neural networks previously trained on a similar problem, which, starting from an image of a tomato leaf, using a transfer learning method, identify if the plant is sick and the type of disease. The proposed networks show a high precision and accuracy coefficient, demonstrating how the application of convolutional neural networks for this type of problem is very effective.