A Novel CNN Architecture for Herb Classification

Herbs are an integral part of any traditional medicine. India is rich in medicinal plants which possess the ability to cure many ailments such as cancer, asthma, arthritis etc. Accuracy in recognition of medicinal plants from a wide variety of plants from forest ecosystems is a difficult task. Convolutional Neural Network (CNN) is one of the deep learning techniques that performs well for many image classification applications. In this paper a CNN architecture has been implemented for herb classification. The proposed CNN consists of four convolutional layers, pooling layers, a fully connected layer and a softmax layer. The ReLU activation function and Adam optimizer improves the performance of proposed CNN with an accuracy of 96.25% with time based decay learning rate.