<![CDATA[Moringa (Moringa oleifera) leaves are widely recognized for their nutritional and medicinal value, making quality assessment crucial in ensuring their market and processing standards. Traditional manual classification of leaf quality is subjective, time-consuming, and prone to inconsistency. This study aims to develop an automated classification system for Moringa leaf quality using a Vision Transformer (ViT) model, a deep learning architecture that leverages self-attention mechanisms for image understanding. The dataset consists of six leaf quality categories (A–F), representing various conditions of color, texture, and defect severity. The ViT model was trained and evaluated using labeled image datasets with standard preprocessing and augmentation techniques to improve robustness. Experimental results show an overall accuracy of 56%, with class-specific performance indicating that the model achieved the highest recall for class D (1.00) and the highest precision for class F (0.74). Despite moderate performance, the results demonstrate the potential of ViT for complex agricultural image classification tasks, highlighting its capability to capture visual patterns in small. Future improvements may include larger datasets, fine-tuning with domain-specific pretraining, and hybrid transformer–CNN architectures to enhance model generalization and accuracy.]]>
Copyrights © 2025
