Classification of healthy and unhealthy banana leaves using deep learning approach: A comparative assessment

India is an agricultural nation where more than half of its population is dependent on agriculture for their livelihood. The lives of the farmers are disrupted by multiple constraints and challenges such as the low yield of crops and occurrence of diseases and pests, soil type, lack of HYV, inadequate irrigation facilities, to name a few. Plant disease is a significant constraint and forms a challenging task. Plant diseases could damage the leaves and fruit and thereby lead to significant economic loss in the agricultural industry. The diseases of the plants are mostly visible on the leaves of the plants and the stem. Banana is an important fruit crop grown in Assam, a state in the North-Eastern part of India. This paper deals with classification of the healthy and diseased banana leaves. The proposed method is based on a hybrid model, a self-constructed CNN model (SCCNN) infused with SVM. The features used for the classification have been extracted using the SCCNN and classified using the Support Vector Machine (SVM) classifier. To verify the robustness and accuracy of the proposed method in this paper, the dataset has been tested with the MobileNet model. The experiments, in this paper, have been performed on the on-field banana leaves images collected from different parts of the state of Assam. The experimental results show that the recognition accuracy of the proposed model is better, achieving a training accuracy of 99.1% and 92.8% testing accuracy, which can realize agricultural applications.