Manufacturers are making strides toward Industry 4.0, a movement to tie a company’s factory floor technology with the internet of things, business and operation systems, supply chain and aftermarket technology, and scores of equipment.
That includes vision inspection systems, which are increasingly going high-tech with the addition of machine learning, artificial intelligence (AI) algorithms that can be trained to catch small blemishes and disfigurements. The information returned from AI inspection systems is part of the massive operating information that can sense, analyze, and respond to changing company conditions.
The resulting data is used to streamline operations and improve efficiency, which lead to massive savings – the premise of Industry 4.0.
AI in computer vision is no stranger to manufacturing Industry 4.0 or to a number of other markets, such as biomedical and consumer goods. The research firm MarketsandMarkets has estimated the AI in computer vision market at $15.9 billion in 2021 and predicted it will grow by more than 25 percent to reach $ 51.3 billion by 2026.
Two years ago, Andrew Ng, among the most prominent figures in AI, stepped into this arena by founding Landing AI, which makes AI vision systems software that can be easily installed and trained using its LandingLens system. As a startup, the company focused on AI inspection for manufacturing systems, but over the years that outlook has grown to encompass other industries, says Kai Yang, vice president of products at Landing AI.
But by its very nature, the tool has not always been easy for factory engineers to deploy on their lines, Yang says.
Deploying deep learning on the production floor
Today, the company announced its LandingEdge, which customers can use to deploy deep-learning based vision inspection to their production floor. The company’s first product, Landing Lens, enables teams, who don’t have to be trained software engineers, to develop deep learning models. LandingEdge extends that capability into deployment, Yang says.
“Strategically, manufacturers start AI with inspection,” Yang said. “They use cameras to repurpose the human looking at the product, which makes inspection more precise.
Ng’s company, like others in the AI vision space, faced a problem: It took an expert to write the code that would integrate the cloud-based platform with a company’s vision system. Getting the image from the factory floor to the cloud so the platform could search for faults — then returning the inspected image back to the factory system — was the purview of a skilled programmer.
LandingEdge attempts to simplify the platform deployment for a manufacturer. Typically users set up a method to “train” their vision system by plugging the LandingEdge app into programmable logical controller and cameras. The PLC continuously monitors the state of cameras and the vision system itself.
After deployment, users present the system with images faults, which, through the help of AI, it gets better and better at identifying, Yang said.
Significantly, AI for vision systems can find dramatically more defects on the factory floor than vision systems that don’t include AI. For instance, an automatic system couldn’t recognize a scratch, Yang said.
“Scratches can have different shapes, depths and color,” he said. “I could write code that says a scratch can be five to 500 pixels in a certain color range, but there’d be no way to enumerate all the possibilities.
“With deep learning, you just label the scratch whenever you see a new one and after a couple of them the system will learn the presentation of the defect,” he added.
AI-driven vision systems competitors
Landing AI isn’t the only maker of AI-driven vision system technology, of course. Kitov.ai and Cognex are two large ones. Fujitsu has also been Fujitsu Laboratories has developed AI-enabled recognition systems for the electronics industry. Many of these companies, like Landing AI, have certification programs that ensure their software is compatible with vision hardware currently on the market.
Last month, for example, Landing AI announced it has joined NVIDIA Metropolis, one such certification program. Many LandingLens customers use the NVIDIA Jetson edge AI platform for their vision hardware system, said Carl Lewis, senior director of customer success at Landing AI.
The program offers opportunities for partners to collaborate with industry-leading experts and other AI-driven organizations, Lewis said.
Ng, among the most prominent figures in AI, regularly extolls the future of moving AI beyond big tech and into manufacturing. While he founded Landing AI to facility 2017 to facilitate the adoption of AI in manufacturing, the company is now seeking to move vision-system AI into other industries as well, Yang said.
The company began with manufacturing because it had been a challenge to send images to the cloud for inspection and to get them back “in a place where a stable internet just doesn’t happen,” he said.
The company is now exploring LandingLens and LandingEdge for use in the health sciences to, for example, watch petri dishes. The system can report when the culture inside the dishes has reached the correct stage for human intervention, Yang said, adding that other uses include the agricultural industry, to ensure plants are healthy and fields are weed free.