Resnet Based Classification in CNN for Ayurvedic Plant Categorization Using Deep Learning

Abstract

There are over a million plant species on the planet. Classifying all these species is very difficult and also results in duplication. Plants can be identified using leaf which carries unique information To extract the various complicated features, leaf images need to be preprocessed. For image categorizing, image classification technique is used. In this study,we combine SVM with ECOC framework (for error correcting) is presented. For identification and classification of objects/images, a new neural network known as Convolutional Neural Network has made more reliable with the performance and computations. CNN is trained with more images with their labels and also used to classify new input images. CNN’s are great at images, we will get new features in objects and are designed to handle huge amount of image data. Here in this study, ResNet50 architecture is used. Our image dataset includes Ayurvedic plant images with 6 categories (classes) and made our classification task a multiclass problem. For multiclass classification problem, ECOC framework is used. One-versus-all coding design is used with SVM classifier. A pretrainedCNN is used for extracting image features. The extracted features are thenusedtoclassifythedata.Theresultfor the class labels are generated by grouping. MATLAB is used to build the program. The experiment yielded a 93% accuracyrate.