Venation Based Plant Leaves Classification Using GoogLeNet and VGG

Abstract

Plants maintain and protect life on earth by perpetuating various seasons and supplying air, food, shade, and shelter. Identification and classification of plants are inevitable for botanists, ayurvedic physicians, ayurvedic medicine manufacturers, and researchers. In this paper, we propose a fine- tuned GoogLeNet CNN model for classification of plants using leaf venation. GoogLeNet and VGG-16 CNN models are trained and tested with Dleaf, Flavia and Leaf1 datasets using CNN and Sopport Vector Machine (SVM) classifier. Fine-tuned GoogLeNet outperforms fine-tuned VGG-16 and Dleaf deep CNN model with a five-fold cross-validation accuracy of 99.2% on Leaf1 dataset using SVM classifier.