Combining CNN and VGG16: A Novel Deep Learning Approach to Onion Purple Blotch Disease Severity Classification

The timely and rationally reliable differentiation of plants by diseases from the perspective of fine agriculture is the most critical element, affecting crop yields and therefore global food security to a very large extent. The proposed research paper will present a stand-alone model that is purely based on 2-dimensional imaging, in which convolutional neural network (CNN)architecture is combined with the VGG16 architecture to classify onion purple blotch disease from five different levels. Combining the advantages of VGG16 and in-depth custom CNN layers for a better and more comprehensive solution will ensure the model's scalable and efficient ability for the early detection and severity classification of this common disease. During stringent training on an exhaustive data set of onion leaf images, annotated to represent different levels of disease severity, the model acquired exceptional classification predicaments. It showed 97.14% overall accuracy with which it made a distinction between phases of the disease. Beyond that, the research not only provides a pathway for automated disease detection in agriculture but also illustrates the possibility of successful applications of a hybrid deep learning model in the process of reduction of the risks stemming from diseases which is one of the aspects of the support for sustainable agricultural practices. The development of future research directions features expansion of the model to application in a wider range of crops and diseases, which are extremely important for coming up with more effective and encompassing precision agriculture advances.