Tobacco Leaf Grading Algorithm Based on Expert ViT Model

Tobacco leaf grading is a crucial step in tobacco production, directly impacting product quality and market value. Traditional grading methods rely on human expertise, which is often inefficient, costly, and highly subjective. With the rapid development of artificial intelligence and big data technologies, large models such as Vision Transformers (ViTs) provide new opportunities for intelligent grading. This paper presents a novel method that integrates deep learning and multimodal data analysis to address the challenges in tobacco leaf grading. Experimental results demonstrate that the improved ViT model achieves an accuracy of 93.2% and a macro-average F1 score of 0.92, outperforming traditional methods by a margin of 14.7 percentage points in accuracy. Furthermore, the system achieves a classification efficiency of 14 images per second, meeting industrial requirements. This study contributes to the field of automated quality control in tobacco production and highlights the potential of large models in industrial applications.