Tomato Leaf Disease Detection and Classification using Deep Learning Techniques

Abstract

This work presents a comprehensive exploration of various methodologies and techniques aimed at improving the detection and classification of diseases affecting tomato plants, a critical aspect of agricultural productivity, particularly in regions such as India, where tomatoes are a staple crop for different use case scenarios. The work encompasses a thorough review of the literature, highlighting the significance of earlydisease detection and the challenges posed by factors such as climate change, soil conditions, and pests. Several deep learning approaches, particularly Convolutional Neural newlineNetworks (CNN), have emerged as promising tools for automating the identification newlineand classification of tomato leaf diseases. This thesis critically analysis and evaluates newlinethe performance of different CNN architectures, including, VGGNet, AlexNet, LeNet, ResNet, and DenseNet in detecting various classes of tomato leaf diseases. Through comparative analyses and experimental findings, the study elucidates the strengths and limitations of each architecture, providing insights into their efficacy in disease detection tasks. Furthermore, the thesis delves into the integration of advanced techniques such as Generative Adversarial Networks (GANs) and Vector AutoregRressive Moving Average processes with eXogenous regressors (VARMAx) to enhance disease identification accuracy and robustness. By leveraging the capabilities of CNNs for feature extraction and classification, coupled with VARMAx mechanisms for improved interpretability and decision-making, the proposed integrated model represents a significant newlineadvancement in disease detection methodologies. Key findings from the thesis include the superiority of certain CNN architectures, such as DenseNet-121, in addressing the vanishing gradient problem and achieved a 98.3% high level of accuracy for disease detection. The study demonstrates the potential of deep learning-based approaches to revolutionize disease detection practices in agriculture, offering proactive solu