Alibaba Cloud GPU Server Helps Defiance Wins 4 Worlds First

MSCOCO (Microsoft Common Objects in Context) is one of the most watched and authoritative games in the world of computer vision. On September 9th, the COCO+Mapillary Object Recognition Joint Challenge Seminar was held as scheduled at ECCV 2018. The competition rankings were announced and the Chinese team won all the championships. Among them, Alibaba Cloud GPU customer – Defiance Technology won 4 world firsts and became the biggest winner of this competition.

It is reported that this year's joint challenge contains a total of six challenges, including COCO4 and Mapillary2. The Defiance Technology team participated in three COCO competitions (example segmentation, panoramic segmentation, human key point detection) and one Mapillary competition (Street View Panorama Segmentation), all winning.

In the field of artificial intelligence, an excellent product is inseparable from the three elements of computing power, algorithms, and data. Behind the success of Vision Technology, it is Alibaba Cloud's latest generation cloud service server GN6 family, which includes the GN6V that has been commercialized and the SCCGN6 that is being beta.

The GN6 family is a server platform designed to meet the high computing needs of artificial intelligence and high-performance computing customers. It has many technological breakthroughs:

GN6 family equipped with NVIDIA Volta architecture Tesla V100 computing card, the maximum specification supports 8 V100 GPU

Each GPU has 640 Tensor cores and 125 TeraFlops deep learning capabilities. Compared to the previous generation GN5 (P100) family, GN6 provides up to 12 times more floating-point computing power in artificial intelligence training applications;

Between the GPUs, with the new generation of NVIDIA NVLink, communicate with each other at speeds up to 300GB/s;

The CPU uses an Intel Xeon Platinum 8163 (Skylake) processor running at 2.5 GHz.

On the storage, it supports Alibaba Cloud Elastic Computing, the latest generation cloud disk product ESSD, which provides ultra-low latency and high reliability cloud disk storage capacity;

Network capabilities, SCCGN6 provides ultra-low latency RDMA networks up to 50Gbps;

In deployment, Alibaba Cloud's GPU cloud service supports one-button CUDA driver auto-installation, with the perfectly supported NVIDIA NGC container image, helping developers to quickly access various optimized deep learning frameworks; GPU host cloud monitoring is supported on operation and maintenance. Real-time understanding of GPU running status;

Alibaba Cloud has a domestic-scale and technology-leading public cloud GPU computing cluster, providing a heterogeneous computing acceleration platform for multiple scenarios, serving more than 1,000 enterprise users. With security, stability, high performance and flexibility, Alibaba Cloud GPU Cloud Servers are helping artificial intelligence, and high-performance computing companies are constantly innovating in technology and succeeding in business!

About Vision Technology:

Defiance Technology is an IoT solution provider with artificial intelligence technology as its core. It is committed to creating maximum value for customers and society with extraordinary technology. As a representative domestic artificial intelligence technology company, it deserves to have nearly 800 patents granted at home and abroad, and participates in the formulation of 19 artificial intelligence national and industry standards on behalf of industry leading technology providers. Including MS COCO, contempt has repeatedly defeated giants such as Google, Facebook, and Microsoft in various international artificial intelligence top competitions, and won the first of 22 world technology evaluations. In terms of commercial landing, contempt for face recognition technology, image recognition technology, intelligent video cloud products, smart sensor products, and intelligent robot products have been widely used in finance, mobile phones, security, logistics, retail and other fields.

About MS COCO:

The full name of MS COCO is Microsoft Common Objects in Context. It originated from Microsoft's COCO dataset, which was funded by Microsoft in 2014. The competition of the same name is the same as the previous famous ImageNet competition. One of the most important and important competitions.