By Arthur Stout
“Overcoming the Challenges of Bringing AI to the Edge” is an insightful new article by Arthur Stout, Director of Product Management – AI Solutions at Teledyne FLIR. The study (a condensed version of which follows) explores how multispectral imaging takes advantage of advanced vision processors, efficient neural networks, large image data sets, and synthetic image data. In particular, he explains how Teledyne FLIR, a subsidiary of Teledyne Technologies, develops powerful artificial intelligence solutions for industrial imaging and defense applications.
Additionally, Stout commented on the article during an interview with Inside Unmanned Systems Editor Abe Peck, describing how these innovations aim to pave the way for ever more accurate modelling.
Specifically, Stout defined three approaches (in-house, open source, and synthetic) to maximize the power of AI:
• Conservator™, a cloud-based data lifecycle management tool developed by Teledyne FLIR to visualize model performance
• ONNX AI, an open source project that establishes a model file format standard and tools to facilitate execution on a wide range of CPU targets
• A collaboration with CVEDIA, a synthetic data technology company, to produce the tools and intellectual property needed to create multispectral data and models using computer-generated imagery (CGI). Teledyne FLIR develops high-performance networks trained on synthetic data in collaboration with CVEDIA and the Teledyne FLIR Research and Development Center.
THE STATE OF AI
“Obviously there are a lot of companies doing good things in autonomous vehicles and such,” Stout said. Inside Unmanned Systems of Goleta, California, where FLIR has a manufacturing and R&D facility. “But I still think we’re just at the beginning, especially in our field, where the economics don’t necessarily align with the investments needed to deploy AI in our end-use markets.”
But there is good news. “I think 2022 will be the breakthrough year, and definitely 2023,” Stout continued. “In the very near future, certainly within the next 12 months, we will be able to offer our customers very sophisticated, very powerful and very precise AI capabilities in these core imaging modules.”
Teledyne FLIR has positioned itself as an “option to buy” for customers. The effort required to embed AI into products and the difficulty low-volume companies face in accessing both high-end vendors and in-depth technical support, Stout said, make the AI’s ability attractive. company to buy silicon and invest in cutting-edge technology at scale.
“Our value proposition is, ‘Hey, we can be your AI source as well as your infrared technology.'”
Flexibility is another attribute of the Teledyne FLIR approach. “Not all applications require the same large networks,” Stout said. “We can reduce frame rate and power and then trade it off for accuracy, object detection, neural network performance metrics, whether it’s frame rate, accuracy or sensitivity, depending on the performance requirements of the customer.
At the same time, Teledyne FLIR needs to support high-end customers with large format images. “Targeted pixels are a very important metric for our clients,” Stout explained. “These automotive customers want to know how far ahead of the vehicle we can accurately detect a pedestrian or vehicle. With automatic emergency braking, we want the widest possible field of vision. Thus, if a pedestrian comes between your vehicles parked on the sidewalk, we can detect it, then control the system to stop the vehicle.
“With AI, at the end of the day, those things never get done. It’s a very iterative process. Patterns are always defined and outliers are always addressed.”