DEEP LEARNING-BASED PLANT DISEASE DETECTION USING MOBILENET V3 AND IMAGE CLASSIFICATION

Abstract

The automation of plant disease identification can significantly enhance agricultural productivity by enabling early intervention. This research proposes a deep learning-based multi-class image classification system using the MobileNet V3 architecture to identify 38 distinct plant disease categories. Using a publicly available image dataset from Kaggle, this study conducted a full model development pipeline including data preprocessing, exploratory data analysis, transfer learning, training with validation, and deployment via a Gradio interface. The system was evaluated using accuracy and multi-class log loss metrics and demonstrates promising results for real-time agricultural applications.