Flower Image Classification Using Convolutional Neural Network
Sandip Desai, C. Godé, Punit Fulzele
Abstract
In the field of pharmaceutical industry, botany and agricultural there is a need of algorithm which will classify the flowers by processing its image. In this context, we propose a flower classification approach based on convolutional neural network. We have applied transfer learning approach for classification of flowers. We have used VGG19 convolution neural network architecture for extraction of features. As we wanted to classify flowers in 17 different classes so we have used 17 neurons in final dense layer of VGG19 convolution neural network architecture with the use of softmax activation function. Results show that we have classified flowers with the validation accuracy of 91.1 % and training accuracy of 100%.