Classification of Plant Leaves of Western Ghats using Deep Learning

Abstract

Countless numbers of plants are available in this world. Identifying each and every plant and then classifying them has become one of the important and difficult tasks.Various parts of plants such as flowers, seeds, leaves can be used for identification, but recognizing leaves is the simplest and most effective method. Deep learning technique brings out effective way of leaf recognition system. Here we have used customised Convolutional Neural Network model to recognize the leaves specially growing in western ghats. A separate dataset has been created by collecting more than 50000 leaf samples of 48 different types of plants. The relevant information about the set of plants are collected from the botanists. Various architectures of CNN such as InceptionV3, MobileNet, VGG16, DensNet are used to evaluate the results. Model gives a satisfactory accuracy of 93.79% on 48 classes.