Classification and Detection of Cabbage Leaf Diseases from Images Using Deep Learning Methods

Abstract

The presented work uses Deep learning methods to detect diseases in cabbage leaves. The plant disease detection is constrained by human visual capabilities. Because, most of the early symptoms detected are microscopic. This process is tedious, time consuming and prediction is a challenging task. Hence, there is a need for developing a methodology that automatically recognizes, classifies and detects plant infection symptoms. Five major types of diseases namely Leaf miner, Diamond backmoth, Blackrot, Maggvots and Downy mildew are considered. Initially, the input images are classified into two types as healthy and diseased. Further, the diseased images are categorized into five different varieties. Around 3000 images of cabbage leaves are used containing healthy and infected leaves. Different phases namely preprocessing, feature extraction, training, testing and classification are used the proposed methodology. The accuracies of 93.5% and 90.5% are achieved for healthy and diseased leaf images. Classification accuracies for different types of diseased images are 89.9, 89.5, 91.8, 90.5 and 90.8 for Leaf miner, Diamond backmoth, Blackrot, Maggvots and Downy mildew respectively. The overall classification accuracy of 92% is attained. The developed methodology is found to provide good classification accuracy. The developed model finds its applications in APMCs, online purchase, Agricultural departments etc.