This study presents a deep learning-based approach for accurate diagnosis of diseases in sugarcane leaves using the Sugarcane Leaf Dataset, which includes 11 different disease classes and 6748 images. The research compares the performance of leading Vision Transformer (ViT) architectures, such as DeiT3-Small and DeiT-Tiny, with popular Convolutional Neural Network (CNN) models. The findings show that there is no direct relationship between model complexity, depth, and accuracy for the 11-class sugarcane dataset. Among the tested models, the DeiT3-Small model achieved the highest performance with 93.79% accuracy, 91.27% precision, and 90.96% F1-score. The study highlights that rapid, accurate, and automatic disease diagnosis systems developed using deep learning techniques can significantly improve sugarcane disease management and contribute to increased yields. The results also indicate that both CNN and ViT-based models can be effectively used in various visual tasks, with each model having its strengths and weaknesses. The study recommends further research to validate the proposed models in live applications and to work with larger datasets for more realistic results.This study presents a deep learning-based approach for accurate diagnosis of diseases in sugarcane leaves using the Sugarcane Leaf Dataset, which includes 11 different disease classes and 6748 images. The research compares the performance of leading Vision Transformer (ViT) architectures, such as DeiT3-Small and DeiT-Tiny, with popular Convolutional Neural Network (CNN) models. The findings show that there is no direct relationship between model complexity, depth, and accuracy for the 11-class sugarcane dataset. Among the tested models, the DeiT3-Small model achieved the highest performance with 93.79% accuracy, 91.27% precision, and 90.96% F1-score. The study highlights that rapid, accurate, and automatic disease diagnosis systems developed using deep learning techniques can significantly improve sugarcane disease management and contribute to increased yields. The results also indicate that both CNN and ViT-based models can be effectively used in various visual tasks, with each model having its strengths and weaknesses. The study recommends further research to validate the proposed models in live applications and to work with larger datasets for more realistic results.