As manufacturers of goods, human inspectors check them for defects. Imagine a scratch on the smartphone glass or a weakness in the raw steel that could work downstream if it is turned into something else. Landing AI, the company founded by Google and Baidu's former AI guru Andrew Ng, wants to use AI technology to identify these errors. Today the company launched a new visual inspection platform called LandingLens.
"We are announcing LandingLens, an end-to-end visual inspection platform that enables manufacturers to create and deploy visual inspection systems [using AI]," Ng told theinformationsuperhighway.
He says the company's goal is to bring AI into manufacturing companies, but he couldn't just repackage what he'd learned at Google and Baidu, partly because it was a different set of consumer use cases partly because there is far less data to work with in a manufacturing environment.
In addition to the level of difficulty, each setting is unique and there is no standard playbook that you can necessarily apply to every vertical. This meant Landing AI needed to develop a general toolkit that any company could use for the unique needs of their manufacturing process.
Ng says in order to put this advanced technology in the hands of these customers and apply AI to visual inspection, his company created a visual interface through which companies can go through a defined process to train models to meet each customer's inspection needs understand.
This is how you take pictures of what a good end product looks like and what a defective product might look like. It is not as easy as it sounds because human experts cannot agree on what constitutes a defect.
The manufacturer creates a so-called error book in which the inspection experts work together to determine what this error looks like based on a picture and to resolve discrepancies if they occur. All of this is done through the LandingLens interface.
Once the inspectors have agreed on a set of labels, they can start iterating a model in the model iteration module, where the company can train and run models to reach an agreed-upon success state where the AI regularly detects the bugs based. As customers run these experiments, the software generates a report on the status of the model, and customers can refine the models as needed based on the information in the report.
According to Ng, his company is trying to introduce sophisticated software to solve a major problem for manufacturing customers. “The bottleneck [for them] is developing the deep learning algorithm, really the machine learning software. You can take the picture and judge whether this part is OK or defective, and our platform will help with that, ”he said.
He believes this technology could ultimately help reshape the way goods are made in the future. “I think deep learning will change the way inspections are conducted, which is really the key step. Inspection is really the last line of defense against manufacturing quality defects. So I'm excited to release this platform to help manufacturers conduct inspections more accurately, ”he said.