Transfer Learning Framework for Precise Multi-Disease Classification in Cauliflower Leaves

Abstract

Early identification of foliar diseases in cauliflower crops is essential to prevent yield losses and support sustainable agriculture. Traditional methods rely on manual visual inspection, which is subjective, timeconsuming, and dependent on expert availability, making large-scale monitoring challenging. With advancements in digital agriculture and Machine Learning (ML), automated disease detection systems offer improved accuracy and faster decision-making. This study proposes Leaf Guard AI, a lightweight and efficient disease detection framework based on a hybrid Inception Residual Network Convolutional Neural Network (IRNCNN) architecture integrated with a Tkinter-based Graphical User Interface (GUI). The system preprocesses input images, extracts hierarchical deep features, and classifies cauliflower leaf diseases into eleven distinct categories with high precision. To evaluate performance, baseline models including Logistic Regression (LR), Decision Tree (DT) Classifier, and Artificial Neural Network (ANN) are implemented for comparison, demonstrating the superiority of the proposed Deep Learning (DL) model. Additionally, the system incorporates a Telegram Bot interface, enabling farmers and agricultural practitioners to upload leaf images via mobile devices. The bot performs realtime analysis and returns classification results along with diagnostic insights, enhancing accessibility and usability. Experimental results show that the proposed model achieves an accuracy exceeding 99%, indicating strong robustness and generalization capability. By combining advanced Deep Learning techniques with user-friendly desktop and mobile interfaces, Leaf Guard AI facilitates automated and real-time disease detection, supporting timely crop management, improved monitoring, and increased agricultural productivity.