Classification of ladies finger plant leaf using deep learning

Abstract

In our nation, the agricultural sector is decreasing which reciprocally affects the production capability of agricultural goods. It has become significantly more important to feed the steadily growing population. Plant diseases affect the livelihood of this imperative source more. It results in depletion of production and economic losses in agriculture. We have to protect and control the crops losses in order to aggregate the agricultural productions. Using agronomists, the traditional method, takes a lot of time to continuously observe the crops and moreover finding an agronomist in rural areas is also very challenging. Therefore, developing a system which can use visible or noticeable symptoms to identify and distinguish plant disease habitually to offer appropriate support will unquestionably help all the farmers to reduce crop losses and will increase production. Currently, the deep learning application in crop disease classification is the most active areas of research for which the prerequisite is an image dataset. A custom CNN architecture is proposed in this article to classify ladies finger plant leaf image into three categories namely healthy, disease and leaf burn. The dataset contains 1088 samples of ladies finger plant leaf images, among which 457 images are identified as healthy (non-disease) leaves, 509 images as disease and pest infected and 122 images as leaf burn due to fertilizer overdose. The images were taken directly from the agriculture farms of different villages of Tiruvannamalai district, Tamil Nadu, India. The proposed CNN architecture achieved 96% classification accuracy.