Industrial AI empowering IoT deployment

“Intelligent video analytics has been a big problem. It’s a big market for us to help our customers.” NVIDIA’s Tom Lin (right) together with Advantech’s Magic Pao discussed next-gen artificial-intelligence approaches to empower IoT concepts.

As the digital transformation of industry matures and the tools, technologies and techniques that fuel transformation grow more widely accepted, it is becoming clear how industrial artificial intelligence (AI) is truly empowering IoT deployment.

“AI is flourishing at the edge to satisfy the massive compute needed for industrial IoT,” explained Grace, the AI voice-generated host for these sessions of the Advantech Connect Online Partner Conference. (How cool is that?!) “AI is a high-performance bridge from local assets to the cloud.”

The responsibility to analyze data and generate insights is increasingly automated, providing sharper guidance and freeing workforces to focus on more strategic roles. This was the central message to the AI-spotlighting presentations during this program of the conference, in which experts dove deep into the changing role of AI from merely a cool new technology to a true competitive strategy enabler for business growth.

Complementing presentations on the role of prebuilt AI in industrial technologies and guidance on managing large-scale AI/IoT edge systems were two sessions with Advantech reps and partners that celebrated real-world applications of AI in the emerging video-analytics space.

Take a look…

AI-enabled computer vision deployments

Magic Pao, senior director of Advantech’s video solution master division, hosted Tom Lin, NVIDIA Taiwan’s vice president of sales, to explore how the partners are contributing to the tremendous progress of AI in the industrial arena—in particular, advanced computer vision powered by edge AI.

“This is particularly true in verticals like smart manufacturing and smart cities,” said Lin, narrowing his focus on new capabilities to get insights from video. “Intelligent video analytics has been a big problem. It’s a big market for us to help our customers.”

Pao explained how AI-enabled video is enhancing performance and lowering manpower needs in agriculture, retail, education, transportation, surveillance and factory settings. But, naturally, difficulties persist. Most-common hurdles include customer disappointment with results, the time commitment to reap results, and the lack among end users of what Pao labels “peripheral technologies” to support smart vision. Example: poor video-streaming setups within a facility. “It’s a real challenge for one company to build all the AI capabilities they need to make this work,” he noted.

The solution? Both Pao and Lin agreed that an ecosystem of solution partners is needed to build a functional AI/computer-vision program and offer customers a speedy path to deployment…a solution kit they can implement before focusing on the last mile of customization. “This is how we can accelerate an edge AI deployment,” said Lin.

Pao detailed the Advantech solutions that make up this pre-built kit—the AI SDK Software that informs the Edge AI System, the AI Pre-Trained Model, and the AI Management Agent.

The duo then explained how NVIDIA solutions are supercharging industrial AI workflows in applications ranging from face-detection to license-plate reading. The Transfer Learning Toolkit, for example, helps develop learning frameworks and models that feed the algorithms that inform industrial processes, all customizable by the end user with usage-domain data that can be retrained as development cycles accelerate.

“In the past, this software development took a lot of time,” surmised Pao. “Now, with pre-trained models we are shortening the product-launch time to market. These tools enable great cooperation and, together, we can grab this AI opportunity with our partners.”

AI with video for operational analytics

Next, Mike Fahrion, CTO for Advantech’s North American IIoT group, hosted Dataperformers’ chief product officer and co-founder, Ali Elawad, to discuss the rapid development of AI-powered video analytics in the operations sector.

“Cameras are the ultimate sensor, providing a rich data source in a physically non-intrusive fashion, which makes them popular in smart cities and smart-space applications, particularly in the manufacturing setting,” said Fahrion, who then noted how the AI solution-provider Dataperformers is leveraging Advantech tools to push the boundaries of artificial intelligence. In short, video analytics is moving out of the lab and into production facilities and products themselves.

“A key factor in this growth is the new wave of GPUs enabling us to train AI algorithms for use in factories,” explained Elawad. “This is realizing the full potential of what we can do here.”

It’s tricky work. Or, at least, it used to be. The pair detailed how visual inspection on a production line, historically hampered by bored, exhausted human eyeballs, evolved into early AI approaches that were better, for sure, but still limited by lighting issues or befuddled by un-fixed objects under surveillance. Rather than optimizing production, these early attempts at automated inspection often led to more recalls, more waste, more hassle than they were worth.

But the current generation of smart inspection, based on detailed audits of customer data and properly trained models, is overcoming past obstacles en route to quicker, smarter processes, reduced scrap, smaller human workforces and quicker adoption.

“Advanced edge computing is critical for this industrial AI,” said Fahrion, who noted that the demand for real-time inspection pushes users away from cloud computing, with its latency and cost restrictions, and closer to the assets themselves. Likewise, the edge-computing approach enables the creation and use of synthetic data, which enhances the inspection process by, for example, building a data set for each defect (particularly valuable when it comes to defects not previously encountered), which boosts accuracy to unprecedented levels.

Added Elawad, “The edge device is the solution to this problem. We work with the Advantech suite of edge devices that helps us deploy our models in factories and guarantee our services. This is enabling the best extent of AI for Industry 4.0. It’s an exciting time.” 

It’s a rapid time, too. Fahrion and Elawad discussed how these AI solutions—which enable human workforces to see their world more clearly—can be launched within a matter of weeks. “For companies in the manufacturing space this will be critical moving forward. The ROI is so compelling,” said the Advantech representative. “This should be a budget item for 2022.”

View the entirety of the Industrial AI Empowers IoT Deployment Track on demand. 

Access the entire agenda of the  2021 Advantech Connect conference.