Potato Diseases Detection Using Image Classification Models in Deep Learning Techniques

Potatoes hold a pivotal role in India's agricultural sector and it is significantly contributing to the nation's economy. Fostering innovation and exhibiting dedication for supporting agriculture fall under the responsibility of the youth generation of this country. Solving the problem of controlling potato diseases is important because it may result huge in crop yield. Consequently, it takes time and human labor to identify and classify potato leaf diseases promptly to reduce the quantum of loss. Technology adoption and routine monitoring have made it possible to detect diseases in their early stages and remove them, thus improving agricultural output. Various deep learning algorithms have been explored to identify diseases in potato plants, including early blight, and late blight, to determine the plant's health by analyzing leaf images. The examined models encompass the Sequential Model, ResNet-50, VGG-16, and Inception V3 each representing a distinct Convolutional Neural Network (CNN) architecture with unique design principles. The evaluation of these architectures and techniques is conducted based on diverse performance criteria. The suggested approach is more robust, since it uses the sparse categorical cross-entropy loss function because the training data are quite balanced. The Neural Networks used in them are densely linked to every other layer to propagate more accurate results of 96.00%, 92.67%, 85.00%, and 94.33% respectively. The algorithms and approaches that have been implemented address and document the effective use of methods for the identification and categorization of diseases in potato leaves. This paper comprehensively explores the visualization of different plant diseases through the application of deep learning models.