Home // CADAT 2024, The First International Conference on Accessible Digital Agriculture Technologies // View article

Variety Classification by Image Recognition of Grape Leaf and Berry

Authors:
Ryo Sugiura
Mitsunori Ishihara
Jihyun Lim
Chisa Iwasaki
Yu Oishi
Hiroshi Yakushiji
Yoshihiko Kuwana
Toshiya Yamamoto

Keywords: image recognition; variety classification; grape leaf; grape berry.

Abstract:
Protection of the rights of plant breeders is essential to preserve the value of new varieties. However, table grapes, for instance, are easy to propagate vegetatively by grafting or cutting, and once they become popular and the cultivation area increases, they face a high risk of unauthorized cultivation and propagation, leading to overseas outflow. To prevent such infringements, effective methods for promptly identifying protected varieties have been desired, and deep learning-based image recognition can be one of the key techniques. The objective of this study is to verify the possibility of identifying grape varieties using images of leaves and berries. The images of leaves, young berries, and mature berries of Shine Muscat and two similar varieties were captured using smartphone cameras, and an image dataset was created to train and test classification models. Image classification models named VGG16, ResNet50, and Vision Transformer (ViT) were applied and redesigned to classify three categories. After training, these models were tested on 51 images of leaves, 174 images of young berries, and 171 images of mature berries. The models achieved classification accuracies of more than 96.1% for leaves, over 99.4% for young berries, and 100% for mature berries. Although additional testing at different sites or in different years will be needed, these results indicate that image recognition techniques can help identify plant varieties toward infringement detection.

Pages: 1 to 6

Copyright: Copyright (c) IARIA, 2024

Publication date: November 17, 2024

Published in: conference

ISBN: 978-1-68558-324-8

Location: Valencia, Spain

Dates: from November 17, 2024 to November 21, 2024