Novel CNN Integration with Pre-Trained model for Enhanced Plant Disease Detection

Abstract

Agriculture is one of the major aspects of national development, yet plant diseases present a significant threat to agricultural production. In order to reduce associated losses and mitigate this threat, plant disease identification in early stages is very essential. Deep learning emerged as a significant advancement in the effective detection of various plant diseases as it can visualize and categorize the symptoms from the leaves of the plants This study examines the state-of-the-art deep learning methods for leaf-based plant disease detection, with a particular emphasis on Convolutional Neural Networks (CNNs), Transfer Learning, and Ensemble Learning. In addition, novel ensemble architecture is proposed, which combines a customized CNN architecture and a retrained model called Google Net. This proposed architecture can detect ten of the most prevalent plant diseases from the Plant Village dataset, including tomato, apple, bell pepper, and potato. The suggested ensemble architecture achieves an astonishing 99.07% accuracy, demonstrating its potential to enhance plant disease diagnostics and promote sustainable agriculture.