Scholarlink

This study introduces an intelligent visual recognition system for 10 indigenous Romanian grapevine cultivars, developed using Convolutional Neural Networks (CNN) with Xception architecture. Images of adult leaves of 10 autochthonous winegrape varieties (Grasă de Cotnari, Fetească albă, Fetească regală, Frâncușa, Zghihară, Tămâioasă românească, Muscat Ottonel, Galbenă de Odobești, Fetească neagră, and Băbească neagră) from the ampelographic collection of the “Ion Ionescu de la Brad” University of Life Sciences in Iași form the training dataset. The CNN is built using Keras and creates a classification system based on the 10 winegrape varieties from the dataset. The training applies a two-phase fine-tuning using pretrained weights using ImageNet. This tool supports precision viticulture and heritage cultivar preservation. This tool is especially valuable for the younger generation, who often use mobile and web applications for learning and research. By combining AI with an easy-to-use interface, it connects traditional viticultural knowledge with modern technology. It encourages interactive learning and raises awareness about Romania’s grapevine heritage in a digital format suited to today’s users.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

AI-Based Recognition Tool for Romanian Indigenous Grapevine Cultivars Using Xception and Convolutional Neural Networks

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • L. C. Colibaba,
  • D. Rîndunică,
  • I. Bodale,
  • R. Butnaru,
  • M. Mustea,
  • L. M. Irimia,
  • C. E. Luchian,
  • M. T. Andrieș,
  • A. Scutarașu,
  • V. V. Cotea,
  • L. Rotaru

摘要

This study introduces an intelligent visual recognition system for 10 indigenous Romanian grapevine cultivars, developed using Convolutional Neural Networks (CNN) with Xception architecture. Images of adult leaves of 10 autochthonous winegrape varieties (Grasă de Cotnari, Fetească albă, Fetească regală, Frâncușa, Zghihară, Tămâioasă românească, Muscat Ottonel, Galbenă de Odobești, Fetească neagră, and Băbească neagră) from the ampelographic collection of the “Ion Ionescu de la Brad” University of Life Sciences in Iași form the training dataset. The CNN is built using Keras and creates a classification system based on the 10 winegrape varieties from the dataset. The training applies a two-phase fine-tuning using pretrained weights using ImageNet. This tool supports precision viticulture and heritage cultivar preservation. This tool is especially valuable for the younger generation, who often use mobile and web applications for learning and research. By combining AI with an easy-to-use interface, it connects traditional viticultural knowledge with modern technology. It encourages interactive learning and raises awareness about Romania’s grapevine heritage in a digital format suited to today’s users.