A Multi-Stage Transfer Learning Framework for Fine-Grained Papaya Leaf Disease Classification using Inception V3 with Adaptive Preprocessing and Robust Field-Condition Generalization

Abstract

Leaf pathogens are very susceptible to papaya crops, among others, Papaya Ringspot Virus, Anthracnose, Powdery Mildew and Leaf Spot lead to a severe decline in yield and quality. These diseases have to be timely and accurately detected to be managed. This study is a proposal to deep learn an automatic disease identification and classification of the papaya leaf diseases based on the image of the papaya leaf using deep learning and the convolutional neural network inception V3. The strategy employs the image preprocessing, augmentation and feature extraction to enhance the recognition process. The model revealed in the results of the experiments are very accurate in classification, competent and robust with a broad spectrum of varied diseases. The system presents an easy and efficient technology to farmers and other agricultural professionals to provide quick-diagnosis and rational decision-making. The approach helps in minimising the loss of crops, enhancing disease control, and sustainable papaya farming.