BEIJING, July 19, 2016 /PRNewswire/ -- As an e-commerce site and buyer-based social network service, www.mogujie.com provides clothes, shoes, fashion accessories and cosmetics for its fashion-forward female consumers - gathering over one hundred million members, to date. When browsing on Mogujie, the user may upload photos of clothes directly on the site. This then prompts the page to reload with products of similar, comparable styles and other related items - making the online shopping experience not only more bespoke and captivating - but also easier for its users. While customers experience this ease of use, the application technology behind the scenes is rather complex.
A powerful, high-performance computing cluster is necessary in order to search for the uploaded photos. Deep learning is the leading application of high-performance computing, and in recent years it has advanced in the capabilities of image identification. Mogujie leverages high speed image identification and analysis by integrating big data with the deep neural network model.
Shortening the computing time and occupying fewer servers is a big challenge for Mogujie. The efficiency level of the traditional server to execute the neural network model is also relatively low. Hence, Mogujie hopes to apply higher-efficiency computing clusters to resolve this challenge.
Inspur's NF5588M3 has changed the traditional single computing unit and coordinates the latest process core of Intel and NVIDIA Tesla/Kepler speed-up computing technology to make CPU and GPU perform their own functions respectively. According to test data statistics released by Mogujie, the node accuracy provided by Inspur high-performance server NF5588M3 improved by double of that provided by the traditional two-way servers with the same computing ability. The performance power consumption cost of the GPU node is also twice of CPU node, which has effectively reduced the direct procurement cost and the usage of space. For all of these reasons, the Inspur NF5588M3 was choosen by Mogujie as the solution to address current challenges, and to be the mainstream model of its community recommendation system.
Inspur's reputable R&D advancements, system construction, operating maintenance service capacity of its super computing products, and software/hardware lines continue to contribute to many of the globe's high-traffic e-commerce sites' demands. Mogujie has become a successful case that has perfectly integrated deep learning technology with Inspur's high-performance products.
Inspur is also currently providing big-scale computing support platforms and facial recognition technologies for internet search giants such as Baidu, Alibaba and Qihoo.
SOURCE Inspur Group Co., Ltd.