Comprehensive study of various machine learning models for plant disease detection: Analysis of deep models on tomato plant

Abstract

Agriculture is a crucial component for an overpopulated country like India, and thus, plant diseases present a substantial risk to the output of crops, thereby making timely identification and diagnosis crucial for guaranteeing robust economic development. Tomatoes, being a prominent agricultural commodity, are vulnerable to a diverse range of illnesses, which can be detected by observing foliage signs. Automated identification, categorization, and assessment of the severity of plant diseases utilizing Artificial Intelligence have become crucial for improving agricultural productivity. Advancements in machine learning, notably in the application of convolutional neural networks (CNNs), provide potential solutions for precisely detecting and categorizing tomato plant illnesses. These automated solutions minimize the requirement for manual inspection, which is both very labor-intensive and susceptible to errors. This work investigates machine learning methods for detecting plant diseases and provides an evaluation of advanced deep learning approaches used for detecting and classifying tomato plant diseases. MobileNet, ResNet, and DenseNet models exhibited greater performance among the six models examined. To improve the interpretability of deep learning models, Grad-CAM (Gradient-weighted Class Activation Mapping) is utilized. The performance of this method is evaluated using high-performing models such as MobileNet, ResNet, and DenseNet, which are commonly used for plant disease detection.