Novel approach for detecting Bacterial spot combining Transfer Learning and Large Language Models

Abstract

Bacterial spot, a pervasive disease affecting Solanaceae crops such as tomatoes and peppers, poses a significant threat to global food security and agricultural productivity. This paper presents a novel approach that integrates transfer learning techniques with Large Language Models (LLMs) to detect bacterial spot with high accuracy and efficiency. We employed state-of-the-art convolutional neural network (CNN) architectures, including Inception-V3 and MobileNet-V2, which were pre-trained on extensive image datasets and fine-tuned on a curated dataset comprising both PlantVillage images and in-house generated photographs from diverse agricultural settings. These images were further enhanced for model training through age and color classification, addressing the phenotypic variability inherent in real-world agricultural environments.The performance of the CNN models was rigorously evaluated, with Inception-V3 achieving a remarkable 97% accuracy in identifying bacterial spot, while the integration with LLMs provided critical insights into disease management strategies, effectively creating a multifaceted decision support system. The LLMs facilitated the interpretation of complex disease symptoms and treatment efficacy reports, enhancing the overall precision of the diagnostic process.Our research underscores the potential of combining transfer learning and LLMs to revolutionize plant disease detection and agricultural practices. This integrative approach not only advances the field of precision agriculture but also opens up new pathways for scalable, data-driven solutions to combat plant diseases and bolster crop health monitoring worldwide.