AI

Edge AI and computer vision technologies are finding new use cases in nearly every industry. In the factory, applications like automated optical inspection (AOI) and industrial robotics improve operational efficiency. In healthcare, these technologies augment medical imaging and diagnostics. And they enable smarter traffic management in our cities, and enhanced security in our offices and public spaces.

This adoption of AI in so many different sectors also changes the outlook of business leaders. AI is no longer seen as some promising technology on the far horizon. AI already is beginning to deliver positive outcomes for organizations of all types. “The implementation of AI in real-world scenarios is currently happening,” says Christine Liu, Product Manager at Advantech, an edge AI platform and AIoT solutions provider. “Decision-makers today view AI as a ‘must-have’ in order to remain competitive.”

It’s a time of great opportunity for AI solutions developers, but they face challenges that need to be overcome, such as adopting AI computing solutions, integrating software SDKs, AI model training, and so on.

The good news is that embedded hardware partnerships enable powerful, development-ready AI computing with products like Advantech’s GPU Card EAI-3101, designed with the Intel^® Arc^™ A380 GPU. GPUs primarily offer visual image optimization and are currently one of the primary AI accelerators used to enhance AI computing power. 

The Latest Embedded GPU Card Supports Multiple AI Use Cases

The EAI series product line offers comprehensive AI acceleration and graphics solutions, including several PCIe and MXM GPU cards with Intel Arc Graphics. With the coming launch of Intel Arc A380E, Advantech offers EAI-3101, a new embedded PCIe GPU card powered by Intel Arc A380E with five-year longevity. Featuring 128 Intel Xe matrix AI engines, this GPU card delivers 5.018 TFLOPS AI computing power. With optimized thermal solutions and auto smart fans, these GPU cards can fulfill different use cases, such as gaming, medical analysis, and more. The designs are proven to outperform the competition in AI inference capability and graphics computing.

The diversity of choices means that OEMs, ODMs, and machine builders are more likely to find a computing platform to suit their needs, regardless of intended use case. Machine builders for the industrial sector, for example, would most likely select one of the commonly used PCIe configuration cards—while the smaller form factor and shock and vibration resistance of the MXM card might appeal to manufacturers of medical devices.

“Intel^® Dynamic Power Share Arc GPUs and Intel CPUs can automatically and dynamically (re)distribute power between processing engines to boost performance depending on the use case—providing stable, high-performance computing for all kinds of edge workloads,” says Liu. “And the Intel^® OpenVINO^™ toolkit helps us accelerate AI inference times, reduce the AI model footprint, and optimize hardware usage.”

Advantech’s #development partnership with @Intel enables the company to bring the latest Intel products to market faster since it has early access to Intel’s latest-generation #processors. @Advantech_USA via @insightdottech

Advantech’s development partnership with Intel enables the company to bring the latest Intel products to market faster since it has early access to Intel’s latest-generation processors. This benefits Advantech customers even when they already have existing solutions in full deployment. For example, ISSD Electronics, a maker of intelligent traffic management solutions, deployed a smart traffic management system in Turkey and recently upgraded the solution to incorporate Advantech’s EAI-3100 series. As a result, the company has already improved its system’s accuracy, reduced AI inferencing latency, and cut construction costs by 33%, says Liu.

Advantech is also announcing new models in its AIR series of edge AI inferencing appliances:

• AIR-150: compact, fanless edge AI inferencing system based on 13th Gen Intel^® Core^™ processors
• AIR-310: edge AI box with MXM-GPU card supported by 14th Gen Intel^® Core^™ processors
• AIR-510: edge AI workstation based on 14th Gen Intel^® Core^™ processors with RTX 6000 Ada

These edge AI platforms adopting the latest Intel platform fit in many different scenarios. Businesses might opt for the relatively lightweight AIR-150 for their offices. To achieve factory AMR automation management, the AIR-310 provides industrial protocols and scalable GPU computing power needed. And for creating a computer vision-assisted medical imaging solution that would likely have heavier graphical computing requirements, the more robust AIR-510 is the right fit.

Leveling the Playing Field for AI Application Development

Alongside its hardware products, Advantech offers a cross-platform edge AI software development kit (SDK). The SDK provides benchmarking tools to evaluate an AI application’s hardware requirements early in the solution development process. This helps developers select the best hardware for their solution—and prevents them from overspending on excessive computing power. In addition, the SDK enables real-time monitoring and over-the-air (OTA) AI model updates post-deployment.

As part of the SDK, OpenVINO provides model optimization and hardware acceleration benefits. The open-source inferencing toolkit also helps AI developers simplify their model deployments and software development workflows by supporting multiple AI model frameworks, including PyTorch, TensorFlow, and PaddlePaddle.

The availability of open-source toolkits and SDKs, coupled with a mature edge AI product ecosystem, will help more machine builders, OEMs, and ODMs to compete more effectively with a stable, development-ready AI computing environment. They help shorten the overall solution development time and allow designers to get innovative products to market faster.

Advantech also offers Edge AI SDK, the AI toolkit, to build an friendly environment from evaluation, SDK adopting, to deployment on all EAI and AIR series products as mentioned above

In the coming years, then, expect to see a far more level playing field for AI application development—what some have called “the democratization of AI.”

In Liu’s view, this is the correct path forward for our increasingly AI-enabled world. “The power of AI shouldn’t be limited to just a few companies. Resources such as our edge computing platforms, our SDK, and OpenVINO are there to be leveraged by everyone,” she says. “AI will be everywhere in the future—which is why we need these open and powerful solutions.”
 

This article was edited by Georganne Benesch, Editorial Director for insight.tech.