by Brian McCarson, Vice President and Senior Principal Engineer at Intel Corporation.
The use of artificial intelligence (AI) for the Industrial Internet of Things (IIoT) is moving into broader adoption. Influenced by open-source communities, the AI domain is making advances quickly – moving from invention to mass replication in just weeks.
To unlock potential AI benefits – discovering valuable insights, realizing efficiencies or gaining competitive advantage – manufacturers need tools that make it easy to perform analytics. Here are key steps to build these AI solutions in industrial settings, using a common use case: machine vision.
Leverage open-source AI innovations
Lighting and camera-resolution inconsistencies are a major hurdle to deploying vision-based analytics on the factory floor. The human eye can correct for different lighting conditions easily, but images collected by a camera naturally vary in intensity and contrast when background lighting varies.
Luckily, the open-source community is fueling democratization and rapid adoption of AI innovations. Members of the community share data, tools and methods that developers can adopt within days. Algorithms are being developed with the ability to absorb lighting variations and neutralize gamma intensity differences as lighting varies throughout the day or as lighting varies from location to location on the factory floor.
Create a closed-loop cycle of innovation
In some machine vision cases, factory managers achieved skewed results from machine to machine because of lighting variations. For example, one machine might achieve high accuracy, low false-positive and false-negative rates while a nearby machine might crash when running the same AI application. Natural light influences from a skylight or dense clusters of light fixtures above a machine might cause these variations.
Fortunately, with the advent of deep learning techniques, algorithms are being developed to replicate the photoreceptors in our eyes and neurons in our brains to absorb lighting variations and neutralize gamma intensity differences.
This is one example of how the open-source community can rapidly deploy solutions to solve manufacturers’ challenges. An example of open-source software is the Intel Edge Insights for Industrial software and the Intel Distribution of OpenVINO toolkit to enable AI in industrial and other sectors. Such free, ready-to-use analytics pipelines were developed for concurrent time-series and video workloads found in industrial applications. They are available in an easy-to-deploy, easy-to-modify, microservices framework. The software supports the acceleration and distribution of analytics on CPUs, GPUs, FPGAs (field programmable gate arrays), VPUs (vision processing units) and Intel Optane memory.
Support open-source communities
The AI market is being driven by exceptionally talented developers and open-source communities. Their innovations are pushing our collective knowledge and capabilities forward, daily. To make these AI innovations and deep learning techniques mainstream, we need to continue sharing tools and resources in ways that are easily accessible, usable, safe, secure, and proven to add value for engineers and managers on the factory floor.
Intel
www.intel.com
Filed Under: AI • machine learning