Classification of Grapevine Leaf Images with Deep Learning Ensemble Models

Abstract

Grape is a fruit with high nutritional value, and its uses are also very extensive, which can be used for brewing and eating. There are many varieties of grapes, and how to identify grape varieties based on the morphology of their leaves is of great significance for grape reproduction and variety evolution. This article takes the leaves of different grape varieties at maturity as the research object. First, the leaf images are collected and preprocessed, and then five pre-trained deep learning models(VGG19, VIT, Inception ResnetV2, DenseNet201, and ResneXt) are trained and fine-tuning using transfer learning. Finally, two voting ensemble machine learning models combine the predictions from the five models. The highest accuracy of 98.1 % is achieved through the ensemble classifier in which decision is made based on soft voting.