Original article can be found here (source): Artificial Intelligence on Medium
Huawei Atlas AI Computing Platform: A Gift for Developers
In Greek mythology, Atlas was a titan condemned to hold up the celestial heavens for eternity after losing the series of battles known as the Titanomachy. Unlike the Greek mythical gods that changed the world with their own power, individuals are flawed, meaning the human race must count on its collective wisdom to thrive. Environments where wisdom is shared have generated great feats such as the computer, the Internet, and artificial intelligence (AI), which have all lifted the human society to a new height. In Greek mythology, Atlas protects the world by holding the sky on his shoulders. Tens of thousands of years later in 2019, the Huawei-developed Atlas AI computing platform was launched, which carries on the mission to push us into the cloud digital era. Let’s take a deeper look at the Huawei Atlas platform and its position in the marketplace. The Huawei Atlas AI computing platform provides diverse product forms — such as modules, accelerator cards, edge stations, servers, and clusters — to help you build an all-scenario AI infrastructure solution across cloud-edge-device. Running on Huawei’s Ascend series AI processors and mainstream heterogeneous computing components, the platform is designed to help supercharge the industries of tomorrow, such as Safe City, Intelligent Transportation, Smart Healthcare, and AI inference. The Atlas AI computing platform provides powerful AI computing for customers to handle massive data volumes. It plays an important role in the Huawei full-stack, all-scenario AI solution. One of the highlights of the Atlas AI computing platform is the collaboration across edge-device-cloud. The cloud is the core of the entire platform for computing and massive data processing. The Atlas 800 AI server provides a high-density, cloud-based AI inference solution that achieves better processing performance with fewer servers. A typical application scenario of the Atlas 800 server is the urban governance system. For example, in a city with a population of over 20 million people and more than 3 million vehicles, approximately 43 million images of passing vehicles will be generated every day. This makes real-time traffic analysis a must for traffic governance. The data analytics of vehicles, traffic violations, and traffic flow requires powerful computing in the cloud. To meet such requirements, either 3,000 servers equipped with general-purpose processors are needed, or 72 to 144 GPU-based servers. However, thanks to the neural processing unit (NPU) processors optimized for AI deep learning, only 60 Atlas 800 servers are required to do the same, greatly simplifying deployment and slashing power consumption. This is the advantage that the NPU-powered Atlas 800 AI server yields over its counterparts in intelligent traffic governance scenario. Edge computing is a supplement to cloud computing. An open platform that integrates the capabilities of network, computing, storage, and application is deployed at the edge to provide services closer to end users. The proprietary Atlas 500 AI edge station and the Atlas 300 AI accelerator card boost the efficiency of edge AI inference, making it an ideal option for industrial quality inspection. Enterprises in the manufacturing industry invest huge manpower in quality inspection. However, due to the intense workload and the dangerous work environments of some special manufacturing processes, the highest accuracy rate of quality inspection can reach only 90% with conventional infrastructure. In recent years, machine-vision quality inspection in the manufacturing industry has increased the accuracy rate to about 95%, but a higher accuracy rate is still required. Many enterprises have purchased expensive equipment for fully automated quality inspection, but human labor is still needed for secondary quality inspection.
Posted on 7wData.be