Crop disease detection with machine learning

Abstract

Plant disease, a significant threat, harms a great deal of farmers each year. A study done by the USDA, (U.S. Department of Agriculture), stated that $220 billion worth of crops are lost each year from plant disease. These diseases can be difficult to discern as the visible symptoms can be overlooked by the human eye. To address this problem, using CNN, (Convolutional Neural Networks), this research identifies crop disease through a dataset of images. The program employs feature extraction techniques using a pre-trained MobileNetV2 model to enhance classification accuracy. After training, the model resulted in 87% testing accuracy. On top of that, a system that recommends certain medication to cure the disease was also implemented. In all, the proposed method displays a way to accurately and efficiently identify plant disease.