Xilinx® Alveo™ Datacenter Accelerator cards are to power a real-time AI-based physical intrusion and theft detection service for SK Telecom – Korea’s largest mobile operator.
SK Telecom’s AI inference accelerator (AIX) implemented on Xilinx Alveo cards provides efficient and accurate physical intrusion detection using deep neural networks. ADT CAPS CO., LTD., South Korea’s second largest physical security company, licenses and commercially deploys the physical intrusion detection service.
SK Telecom’s AI-based physical intrusion detection service, T view™, monitors hundreds of thousands of customers’ commercial and home camera systems in real-time and dispatches security guards under physical intrusion circumstances. Processing a large volume of data from thousands of cameras using deep neural networks necessitates a powerful AI accelerator that can provide enough throughput and accuracy.
T view uses SK Telecom’s AI inference accelerator (AIX), implemented on Xilinx Alveo U250 cards. Running on servers in SK Telecom’s data centre, Alveo U250 cards have demonstrated high throughput and accuracy in theft detection services. The cards deliver lower detection latency and have achieved superior throughput (frames per second) compared to leading GPUs and have proven to be more cost effective. The improved throughput and accuracy delivered by Alveo cards means more customers will have access to real-time, AI-based security services, and a more reliable protection against security threats.
“In the era of Artificial Intelligence where new services are being deployed at unprecedented rates, we keep pursuing to innovate our cloud systems to deliver more value to our customers with more reliable and efficient services across diverse segments,” said Park Jin-hyo, Chief Technology Officer and Head of ICT R&D Centre of SK Telecom.
He added: “AIX in collaboration with Xilinx represents one of our efforts to streamline the cloud system for AI services.”
Xilinx Alveo accelerator cards are designed to meet the performance and flexibility needs of data centre AI workloads, providing 10X higher performance1 for AI-based speech translation and over 3X higher throughout2 for video analytics pipelines compared to GPUs.
Built on the Xilinx® 16nm UltraScale+™ architecture, Alveo accelerators are adaptable to changing algorithms and acceleration requirements, enabling domain specific architectures that optimize performance for a range of workloads without changing hardware, and while reducing overall cost of ownership.
The commercial deployment of Xilinx Alveo cards in SK Telecom’s data center for AI services builds upon the announcement and success of SK Telecom’s automated speech-recognition (ASR) application powered by Xilinx FPGAs to accelerate NUGU, a voice-activated assistant.
“This represents an important win in the commercial AI services space and builds upon our collaboration with SK Telecom to significantly improve theft detection and prevention, where our leadership in real-time compute and analysis helps to promote AI-based services to the South Korea market,” said Donna Yasay, vice president of marketing, Data Center Group, at Xilinx.
“The adaptability, low latency and high throughput delivered by our Alveo accelerators makes them ideally suited for the visual AI inference workloads in the T view service.”
SK Telecom’s AIX accelerator, implemented on the Alveo U250, and the T view service are currently available in South Korea and are expected to be available globally in fall 2020. The T view is also expected to be commercially available as a rebranded solution from other SK Telecom licensed companies.
- Transformer NMT (symbols/sec) benchmark of Nvidia Tesla T4 GPU and Alveo U50
- Full video analytics pipeline including decode and CNN inference – compared to NVIDIA Tesla T4 vs Alveo U200 of 1080p60 video streams