Leveraging MobileNetV3 for Advanced Grapevine Leaf Classification in Viticulture

Using artificial vision approaches, we address the significant challenge of grapevine species recognition from leaf photos in this study; this is a necessary task in getting sustainable viticulture by enhancing crop management, disease detection, and production optimisation. Accurate species identification of grapevine ensures food security and supports responsible consumption and production techniques, by means of which it guarantees sustainable agriculture. Leveraging the knowledge acquired from previous tasks, we approach this difficulty using pre-trained MobileNetV3Large architecture and transfer learning to increase grapevine leaf categorisation accuracy. Five grapevine species—"Buzgulu," "Ala_Idris," "Nazli," "Dimnit," and "Ak"—represent 500 photos in our collection. We split the data in 80% for training (400 images) and 20% for validation (100 images) to evaluate the model strictly. This approach produced amazing 96% accuracy in identifying the grapevine species, therefore highlighting the prospects of sophisticated artificial vision technology in support of environmentally friendly farming methods. Particularly in relation to zero hunger, ethical consumption and production, and life on land, this study advances more general objectives of sustainable development.