Abstract :

Tobacco Crop Productivity Is Significantly Affected By Leaf Diseases, Which Often Remain Undetected At Early Stages Due To Strong Visual Similarity Among Symptoms. Automated Disease Identification Can Assist Farmers In Timely Diagnosis And Effective Disease Management. This Paper Proposes An End-toend Tobacco Leaf Disease Detection And Advisory Framework That Integrates Deep Learning–based Visual Analysis With Knowledgedriven Information Retrieval. Discriminative Information Enhanced Contrastive Vision Transformer (DIEC-ViT) Is Employed For Disease Classification Using A Tobacco-specific Leaf Image Dataset. The Proposed Model Achieves A Test Accuracy Of 88.61%, Outperforming Baseline Convolutional Neural Network (CNN), ResNet, And Recurrent Neural Network (RNN) Models In Terms Of Accuracy, Convergence Stability, And Generalization Performance. To Enhance Practical Usability, A Structured Tobacco Disease Knowledge Base Is Semantically Encoded Using Sentence Embedding Algorithms And Indexed With FAISS For Efficient Similaritybased Retrieval. A Retrieval-augmented Generation (RAG) Mechanism Is Further Incorporated To Ensure That Advisory Responses Remain Factual, Reliable, And Explainable. Experimental Results Demonstrate That The Integration Of Transformer-based Visual Modeling With Semantic Retrieval Significantly Improves Both Disease Recognition And Decision-support Capability. The Proposed Framework Offers A Robust And Deployable Solution For Intelligent Tobacco Disease Management In Real-world Agricultural Settings. Index Terms—Vision Transformer, Tobacco Leaf Disease Detection, Plant Disease Classification, Retrieval-Augmented Generation, FAISS, Precision Agriculture

Published:

25-3-2026

Issue:

Vol. 26 No. 3 (2026)

Page Nos:

684-692

Section:

Articles

License:

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

How to Cite