Deep Learning Based Image Cognition Platform for IoT Applications

Abstract

With more and more Internet of things(IoT) applications are developed over the website, image retrieval is widely used in Web applications. Usually for image retrieval, the queries are needed to be presented in literal language and the retrieved images are captioned in advance. Deep learning is a recent solution to image retrieval, which could calculate the similarity of two images by extracting their features without captions. In this paper, an image cognition platform based on deep neural network is proposed to address the problem of image retrieval as well as image caption. The platform combines the convolutional neural network, recurrent neural network and hash technology together to provide interfaces for users to query with both images and natural language. A case study is given to verify our platform. The result shows that our platform can help managers to efficiently browse similar services provided by other auto 4S stores through image retrieval.