Plant Disease Detection Using SqueezeNet and MobileNetV3

The identification of plant diseases is essential for reducing crop loss and promoting sustainable agriculture. This study centers on the categorization of plant diseases with deep learning, namely the lightweight SqueezeNet model, selected for its efficiency and appropriateness for implementation on resource- limited platforms. A bespoke dataset consisting of leaf pictures from five plant species—Tomato, Potato, Pepper, Rice, and Millet—encompassing 19 disease categories was assembled and subjected to preprocessing via image scaling and normalization. The SqueezeNet model was trained and evaluated against other CNN architectures such as MobileNetV3 and ResNet50. Among them, SqueezeNet achieved competitive accuracy while keeping a substantially reduced parameter count, making it appropriate for realworld applications. This project's ultimate objective is to create a web application that enables to use real-time diagnosis of plant diseases using leaf photos. In order to facilitate prompt and well-informed crop management decisions, this solution seeks to provide farmers and agricultural professionals with an easily accessible, precise, and quick diagnostic tool.