Detection of Some Grape Varieties with Deep Learning Techniques




Deep learning, CNN, Classification, Grape, Ampelography


While determining grape varieties in viticulture, characterization features of shoots, leaves, clusters and fruits are used. These characterization features, in order to form an international method union, with a working team formed on behalf of the "International Board for Plant Genetic Resources" and the "Office International de la Vigne et du Vin-OIV" and "International Board for Plant Genetic Resources". It has been published in a norm called 'Grape Descriptors', developed in collaboration with the international union for the Protection of New Varietes of Plants-UPOV. The ampelographic characteristics of grape varieties are determined according to the characterization features in this norm. Each grape variety has ampelographic characteristics specific to its shoot, leaf, inflorescence and fruit. After these features are determined according to the grape descriptor norm, they are expressed numerically or verbally. In this study, using ampelographic features, Corint, Merlot, Tayfi, Michele palieri, Narince grape varieties were classified using deep learning techniques. The aim is to determine which grape variety it is by using the ampelographic characteristics of grape varieties with deep learning techniques. A new CNN model consisting of 15 layers was created for the study A total of 1028 images of 227×227×3 dimensions of the clusters and fruits of five grape varieties were used in the data set with five classes. 80% of the images are reserved for training and 20% for validation. In the MATLAB program, 96.10% classification success rate was achieved with the new and originally developed CNN model. It was concluded that the CNN model created as a result of the analyzes was successful and could be used in the classification of grape varieties.



How to Cite

Terzi, İsmail, Özgüven, M. M., & Yağcı, A. (2023). Detection of Some Grape Varieties with Deep Learning Techniques. Turkish Journal of Agriculture - Food Science and Technology, 11(1), 125–130.



Research Paper