TeraDeep's Industry-First FPGA-based AI Inference Fabric Speeds Image Recognition, Video Analytics for On-Premise Appliances
/EINPresswire.com/ -- Machine Learning Start-up Taps Xilinx FPGAs, Micron Hybrid Memory Cube Technology to Achieve Industry-Best Low Latency Performance
SANTA CLARA, CA--(Marketwired - Oct 18, 2016) - TeraDeep, a developer of accelerated deep learning applications and AI appliances, has completed the development of a ground-breaking product for fast-response deep learning applications, including on-premise image recognition and video analytics for high-value asset protection.
For low-latency applications such as video analytics, where quick recognition and tracking of fast-moving objects is critical, the graphical processing units (GPUs) used in conventional systems are at a disadvantage. TeraDeep instead uses an FPGA-based architecture that offers faster analytics at half the power, making it an ideal candidate for on-premise appliances.
The first version of the company's solution is an FPGA-based PCIe board that achieves a four-time lower latency compared with the latest GPUs. The board, which runs TeraDeep's deep learning acceleration technology, TD Accel™, is a key component in the company's first AI appliance. AI appliances are expected to help bring about the development of "AI Cities," an extension of "Smart Cities."
"We found that our technologies are a highly effective match for on-premises appliances used in security and safety applications," said Didier Lacroix, TeraDeep's CEO. "We are delighted at the level of performance our close partners Xilinx and Micron have delivered to our first board. With our new market focus, we expect our collaboration to intensify."
Expertise in Accelerating Deep Learning
TeraDeep has developed a suite of advanced deep learning algorithms, optimized to run on the company's TD Accel™ acceleration platform. The fourth generation of TD Accel™ was developed over the past year and is the result of the close collaboration with Xilinx and Micron.
Since obtaining funding from Xilinx at the end of last year, the two companies are actively cooperating to facilitate the development of highly efficient embodiments of the TD Accel™ technology. The collaboration culminated in several FPGA board architectures and the announcement of the first board today. Other versions will be announced for a wider range of FPGAs in the near future.
The collaboration with Micron extended to the novel application of the company's FPGA-based computing platform and the Hybrid Memory Cube (HMC). In particular, HMC technology enables efficient, high-bandwidth access to random sets of pixels. This access boosts video processing significantly, especially when using convolutional neural network algorithms.
"Capitalizing on memory bandwidth is critical to unlocking next-generation computing solutions," said Steve Pawlowski, vice president of advanced computing solutions at Micron. "TeraDeep's high-performance deep learning solutions demonstrate the kinds of innovation that can spring from the combination of advanced memory silicon and clever engineering."
A demonstration of TeraDeep's AI solution running on Micron's Pico Computing AC-510 accelerator boards with HMCs and Xilinx Kintex Ultrascale FPGAs will be on display at Micron's booth at the Supercomputing 2016 conference November 14-17 in Salt Lake City, Utah.
TeraDeep is a pioneer in the development of specialized neural networks and acceleration technology for deep learning. The company develops and markets unique AI Appliances offering complete solutions embodying a creative combination of specialized hardware designed for purpose-trained networks, available through a robust set of APIs. TeraDeep AI Appliances are leased either directly or through its value-added partners. It licenses reference designs and pursues OEM models with qualified parties. For more information, visit www.teradeep.com.
For more information contact: