May 31 2018
Developing world's fastest Neural Network Inference Engine to accelerate native deep learning inference applications in next-generation Edge AI devices
VIA Technologies, Inc., today showcased its family of high-performance Edge AI systems for the Automotive, Enterprise IoT, and Smart City segments at its global headquarters and highlighted its core facial and object recognition, 360-degree surround view, and ADAS (Advanced Driver Assistance System) technology platforms that are accelerating the development and deployment of groundbreaking real-world applications by customers in world markets.
"Edge AI systems that can see, hear, understand, and respond to their surrounding environment in real-time without having to rely on connections to the cloud provide the key to boosting innovation and creating engaging new customer services and experiences in industries as diverse as transportation, retail, entertainment, and medical," commented Epan Wu, General Manager of the VIA Embedded Division. "With their scalable, high-performance designs, VIA Edge AI Systems will unlock the unlimited potential of artificial intelligence for transforming the world we live in."
VIA Edge AI systems are powered by best-of-breed SoC platforms, including the Qualcomm® Snapdragon™ 820E Embedded Platform and NVIDIA Jetson TX2 module, and deliver the optimum blend of compute, graphics, and connectivity for the target AI application scenarios. System functionality and performance can be further optimized with the integration of VIA Smart Facial Recognition, VIA Mobile 360 ADAS, and VIA Mobile360 Surround View Computer Vision technology platforms that enable a diverse array of intelligent video AI applications including:
•Emotion, age and gender detection and people counting and tracking for security, surveillance, and customer engagement applications
•Real-time 360° in-vehicle monitoring and recording for up to six individual HD camera streams for tracking driver behavior and boosting awareness of road conditions
•Lane Departure Warning, Forward Collision Warning, Blind Spot Detection, Pedestrian Detection, Vehicle Detection, Speed Limit Detection, and Rear-end Collision Avoidance for enhancing vehicle safety
To enable the integration of native deep neural network processing capabilities in next-generation Edge AI servers, systems, and devices, VIA has initiated the development of the world's fastest Neural Network Inferencing Engine. In addition to support for 8b & 16b quantized-integer inference compatible with Google's 8b TensorFlow Lite, this high-performance engine also features 16MB SRAM internal memory, 16 TB/s internal bandwidth, and 16 TOP/s plus a host of architectural optimizations for real-world AI applications.
"The VIA Neural Network Inferencing Engine promises to provide the most powerful solution for accelerating deep learning inference applications like image recognition, speech recognition, and natural language processing," concluded Wu. "With its unmatched performance, it will drive the emergence of new generation of Edge AI devices that will inspire a wave innovative new applications and services that will transform the world we live in."