DETECTION OF PLANT LEAF DISEASES USING DEEP LEARNING CONVOLUTIONAL NEURAL NETWORKS

Abstract

The maturity of the machine vision and machine learning models has enabled it to create automated software to achieve classification. The simplicity of computation and minimization of operator intervention form the basis for using these methods in surveillance units for security. Therefore, machine learning, computer vision, and image processing are now assisting every way of life. In this study, a convolutional neural networks (CNN) based technique has been used to learn features from cotton leave diseases, i.e., whitefly, Jassid, Thrips and Lashkri Sundi, and healthy (normal). The dataset has been collected locally, consisting of 576 images of whitefly, 2216 images of Jassid, 1504 images of Thrips, 568 images of Lashkri Sundi, and 686 images of normal. The proposed technique auto performs feature engineering on the images. Experiments have been performed on each of the red, green, blue, and intensity channels, as well as on whole RGB dataset with 5-fold cross validation. Results show that the proposed model achieved an overall accuracy 86.57% (on average).