Embedded World 2019: Optimising AI with computing at the edge
ADLINK Technology is delivering heterogeneous computing platforms optimised for Artificial Intelligence (AI) at the Edge to quickly turn data into action at the point of generation. ADLINK has been developing AI products in collaboration with chipmakers NVIDIA and Intel and has formulated a hardware optimisation strategy to help system architects address performance, Size, Weight, and Power (SWaP) requirements for Edge Computing systems.
The company invites attendees of the upcoming embedded world 2019 event in Nuremberg, Germany to explore ADLINK’s demonstrations of Edge AI applications in Hall 1, Stand 540 from 26th to 28th February 2019.
“ADLINK is taking embedded computing to the next level with GPU- and VPU-accelerated heterogeneous computing platforms. As an NVIDIA Quadro Embedded Partner with extensive experience in embedded applications, ADLINK is uniquely qualified to bring NVIDIA Quadro GPUs with GPUDirect, NVIDIA Video Codec SDK, custom firmware and longevity support into embedded applications, enabling them to tap the power of embedded graphics and AI,” said Edgar Chen, Managing Director of ADLINK.
ADLINK’s heterogeneous computing platforms integrate more than one kind of processor or core to process complex images and data and real time AI for embedded applications including:
Industrial Imaging: Multiple, high resolution displays required for air traffic control, Electronic Chart Displays and Information Systems (ECDIS), video walls, digital signage, gaming, and healthcare.
Parallel Computing: High performance application processing for radar/sonar systems in military and aerospace, ultrasound imaging in healthcare, and accelerated multi-access Edge Computing (AMEC) in telecom.
AI Engines: AI training and inferencing in smart manufacturing, smart city, telecom, defense, and transportation.
As previously highlighted, a series of AI applications at the Edge will be demonstrated at ADLINK’s Stand 540, Hall 1 at embedded world 2019, including:
- An Automated Optical Inspection (AOI) application, based on ADLINK’s next generation, fanless, embedded computer MVP-5100 and embedded graphics module EGX-MXM-P1000, will find solder defects on printed circuit boards in varying illumination conditions.
- An automated meter reading system equipped with the deep learning accelerator EDL-mPCIe-MA2485 will recognise and collect readings from mechanical metres.
- A wood inspection application will classify samples by type with the AI-enabled smart camera NEON-J and look for wood scratches using AI-enabled vision platform EOS-J.
- AI-accelerated frame grabbers PCIe-GIEIMX and PCIe-GIENVQ will detect cracked candies and read deformed labels.
ADLINK heterogeneous computing platforms consist of GPU- and VPU-accelerated board-, system-, and server-level products, enabling system architects to construct and optimise system architecture for both AI inferencing and training applications.
With intelligence moving to the Edge, ADLINK heterogeneous computing platforms perform real time streaming of data between Edge devices and systems, ultimately leading to better decision-making.
In addition to its large variety of heterogeneous computing products, ADLINK offers consultancy services via deep learning profiling to help users determine the right platform to cost effectively fulfill their applications needs. ADLINK is also working with research bodies and academic institutions to find bottlenecks on AI platforms to profile system issues that can be addressed for performance improvement.
Visit the ADLINK stand for a close look at how AI works at the Edge.