Grape Leaf Disease Detection and Classification Using Machine Learning

Grapevine diseases and pets could cause significant financial losses to grape production and farmers if not detected and treated early. With the recent advance of artificial intelligence techniques and machine learning technologies, people start to use computer vision and deep learning algorithms to detect and classify grapevine diseases with high efficiency. In this paper, four modified deep learning models are developed for grape leaf disease detection and classification based on a developed grape leaf dataset. Transfer learning technique has been used in this research project based on three pre-trained machine learning models (VGG16, MobileNet, and AlexNet). The targeted diseases include: black rot, black measles, leaf blight and phylloxera. The reported comparative evaluation results show better accuracy and performance improvement compared with pre-trained models. In addition, an ensemble model based on these four developed models improves final detection and classification accuracy. The reported evaluation results show a great potential usage in grapevine production.