Millet Blast and Leaf Smut Disease Prediction Using CNN

Abstract

This research paper shows a deep learning-based image classification system of automated millet disease based on convolutional neural networks. The model as proposed is the field level image of millet leaf reduced to a standard dimension and smoothed so that the features obtained are similar across all the images. The CNN structure features a series of convolutional, pooling, flattening, and fully connected layers along with their former embedded in it that will be trained to acquire hierarchical visual patterns that are distinctive of various categories of diseases. The data is coded into several categories, and the labels are represented in the categorical coded form. The model itself is trained corrupting the Adam optimizer and categorical cross-entropy loss and tested on unknown test data to determine its ability to generalize. The suggested solution is expected to assist in the early detection of diseases in the millet crops so that nuances of the inspection work can be minimized and the agricultural intervention can be performed at the necessary time. The system shows how computer vision techniques can be applicable in monitoring overall smart crop health and sustainable precision agriculture applications.