Socionext has teamed up with Hailo Technologies, an innovative AI chipmaker, to launch a next-generation processing solution to deliver AI video analytics at the edge. The system combines the Socionext SynQuacer™ SC2A11 multicore SoC optimized for low-power edge devices with the Hailo-8™ deep learning processor.
The combined edge-based solution is a highly scalable and efficient platform enabling top performance for AI-enhanced video processing and analytics. Foxconn, with its BOXiedge edge computing solution, is one of the first companies to deploy this game-changing technology.
Read on to learn more.
Traditional Cloud Servers Bottleneck Smart Technologies
First, a few comments on why the future of smart technologies resides at the edge and not in the cloud.
Sensor-enabled technology is at the heart of exciting innovation in the retail, smart city, and industrial sectors. However, processing and analyzing all this data, especially from video sources, remains a significant bottleneck. The traditional cloud server systems in widespread use today cost too much, use too much energy, and take up too much space.
As we have written about before, further advancements in these and other smart technologies require more affordable and efficient solutions for video processing and analytics – and these systems need to be at the edge.
Why the edge? Consider emerging applications in smart retail, smart cities, and industrial IoT that utilize a large number of cameras to do things like monitor in-store activity, ensure production quality, or manage urban traffic. For this information to be valuable, it must be processed quickly, efficiently, and with minimal latency, which is not practical in the cloud. Using traditional cloud servers is no longer viable to scale smart applications cost-effectively.
Overcoming this bottleneck is why we introduced our multicore SynQuacer line of SoCs. The SynQuacer SoCs are cost-effective and energy-efficient chips capable of delivering the multiprocessing required for robust AI-powered video analytics at the edge.
Edge Video Analytics Powered by Socionext SynQuacer SC2A11 & Hailo-8
Our collaboration with Hailo Technologies, combining the Socionext SynQuacer SoC with the Hailo-8 deep learning processor, illustrates the performance and efficiency advantages of an edge-based AI processing solution.
Socionext SynQuacer SC2A11
The Socionext SynQuacer SC2A11 incorporates twenty-four ultra-low-power ARM CortexTM-A53 cores operating at 1 GHz while supporting up to 64 GB of DDR4-2133 ECC memory. The device is an ideal foundation for low-cost, highly-integrated, and power-efficient server systems deployed to handle edge computing, Internet of Things (IoT) data processing, and cloud service applications.
SynQuacer SC2A11
Hailo-8 AI Processor
Measuring at only 15 x 15mm, the small size Hailo-8 AI chip features an innovative architecture capable of performing up to 26 Tera Operations Per Second (TOPS), a level of performance that enables edge devices to run sophisticated deep learning applications that could previously run only on the cloud. The advanced structure of the Hailo-8 translates into higher performance, lower power (< 2W under typical operation), and minimal latency.
Hailo-8 AI Processor
Use Case Example – Foxconn BOXiedge Edge Server
Foxconn BOXiedge
The Foxconn BOXiedge edge computing solution is one of the first application examples to come out of our collaboration with Hailo Technologies. Under the hood, the BOXiedge device features the SynQuacer SC2A11 high-efficiency parallel processing SoC integrated with a Hailo-8 AI M.2 module. A demonstration of this was featured at CES 2020. Follow the links to view the demo brief and Hailo’s video.
The BOXiedge system, capable of processing and analyzing input from over twenty cameras in real-time, is a cost-effective and scalable smart video management system (VMS) with advanced AI video analytics features.
Sample Edge AI Computing System Featuring:
- Multiple surveillance cameras connectivity with existing ethernet cable
- Connects up to 10-20 ch cameras with real time AI processing at a typical 35W power consumption
- Easy data upload and download with the updated AI model
Diagram of the Edge AI Computing System
Sample Edge AI Computing System Featuring:
- Object/person detection
- Image classification
- Range determination
- Position tracking
- Facial recognition
- Pose estimation
- Object/person detection
- Image classification
- Range determination
- Position tracking
- Facial recognition
- Pose estimation
Enabling Advanced Edge Solutions for Smart Cities, Smart Retail and Industrial IoT
To sum up, realizing the full potential of smart applications, especially those involving video, requires quick processing at the edge. Having data processed locally using smart devices translates into significant cost reductions and energy efficiency. It also offers faster processing, which directly translates into more actionable insight and value from video data.
For vendors looking to become tomorrow’s leaders in smart technologies involving video analytics, Socionext is here to be your SoC hardware partner. Contact us today to start the conversation about how we can help you create high-performance, energy-efficient, and scalable AI edge solutions.
