Research on Flower Retrieval Based on Deep Learning

Abstract

Traditional flower retrieval system uses the technology of the low-level visual feature extraction and image similarity measurement, which has poor generalization ability and low retrieval efficiency. In order to obtain more detailed and abundant image features, a method of flower feature extraction based on deep convolution network is proposed. The deep learning model of VGGNet convolution neural network is used to realize flower retrieval. The experimental results of Oxford 102 flower data set show that the method based on VGG16 model has the characteristics of high accuracy, fast query speed and good robustness.