By combining evolving cameras with faster and smarter automatic object recognition capabilities, computer vision has become one of the most promising applications of AI in the future.
At the Transform 2020 digital summit, Intel Vice President Brian McCarson Matt Marshall (Matt Marshall) of VentureBeat is the CEO of the rapidly growing Industrial Internet of Things (IIoT). The role of vision in the marketplace was presented.
In this conversation, we’ll focus on the new use case we have to help with: significantly improved product defect detection, which is expected to improve the reliability of everything from computer screens to automobiles. I put it in
McCarson said that historically, manufacturers working to eliminate product defects did not lack staff or defect detection experts. In other words, the limitation of the human eye is a hindrance.
Defects in today’s consumer products can be microscopic, such as screen pixel defects and surface defects in aluminum components in automobile gearboxes.
Humans are good at detecting changes in movement and style, but we can’t always find those nuances. Therefore, with the development of computer vision, Intel sees an opportunity.
Intel has developed a computer vision solution in cooperation with cloud service provider Alibaba. This solution increased the positive defect detection rate of vehicle metal workers from approximately 20% to over 99%. In short, the difference is that 4 out of 5 defects and 1 out of 100 defects are missed.
At least for the computer vision sensing components, this is a noticeable change, greatly improving the reliability of the car.
McCarson noted that the solution is reasonably priced and “scalable.” This is because manufacturers can add this solution to a million dollar production line for just a few thousand dollars without making other manufacturing changes.
He said, “So after several days of real return on investment (ROI) measurement, manufacturers can implement one of the most prestigious defect quality control measures in the world.
” Intel is currently working with hundreds of companies with similar Factory Cooperation for quality control. Factories that implement this quality control can improve production yields, reduce yields, and increase operating margins, while reducing negative manufacturing and environment-related yields.
McCarson explained that another great advantage of IIoT is the use of open source software. Open source software can help companies without extensive AI experience develop high-performance machine vision solutions.
For example, Intel’s free convolutional neural network toolkit, OpenVINO, comes with a ready-to-use visual inference model that has been applied to various use cases and is 80-90% performing out of the box.
We have effective machine vision solutions that can help you achieve this. By tuning this solution, the company can achieve 96% to 97% performance.
McCarson suggested that open source has become part of the industry trend, moving from garden devices with proprietary walls (like BlackBerry) to platforms with a foundation for future innovation.
Given the nature of Intel’s artificial intelligence solutions (more flexible than specific uses), the company considers providing customers with a guarantee for the future, rather than simply buying enough products for today’s applications. , I want to buy flexibility to adapt to future AI needs.
When Marshall asked about the market opportunities for computer vision, McCarson said it was “very surprising.” In the consumer goods manufacturing industry, the annual global turnover for industrial automation alone amounts to US $ 500 billion, in order to effectively solve manufacturing problems that were almost impossible to solve a few years ago.
We already know that we are ready to embrace AI and computer vision. Over the next two to five years, edge-based computer vision and data analytics will grow simultaneously to improve defect detection, inventory tracking, and machine uptime, including data analytics. of traditional time series and modern computer vision. He predicts that there will be a “big change” in usage.
At the summit question and answer session, McCarson was asked about the future trend of AI hardware. He replied that one of the main issues is to provide hardware that can adapt to future changes.
The company now wants to use software to update its hardware seamlessly, rather than relying on the old approach of having to submit tracks for updates.
He also pointed out that the current stage of AI is limited by the availability of the model when performing certain tasks, rather than hardware performance limitations. He also recommends that software developers be responsible for creating models that take full advantage of available technology.
Odyssey has been the lead content writer and content marketer. He has vast experience in the field of writing. His SEO strategies help businesses to gain maximum traffic and success.