Computer vision specialist Landing AI has a unique calling card: its co-founder and CEO is a tech rock star.

At Google Brain, Andrew Ng became famous for showing how deep learning could recognize cats in a sea of ​​images with astonishing speed and accuracy. Later, he founded Coursera, where his machine learning courses have attracted nearly five million students.

Today, Ng is best known for his views on data-centric AI – that improving AI performance now requires focusing more on datasets and less on refining models of data. neural networks. It’s a philosophy encoded in Landing AI’s flagship product, landing target.

Founded in 2017, Landing AI counts Foxconn, StanleyBlack&Decker and automotive supplier Denso among its users. They and others have applied deep learning to improve efficiency and reduce costs.

A ranking challenge

A chipmaker with manufacturing plants around the world was one of the first to try LandingLens. It wanted to use deep learning to improve the throughput and yield of the wafers that transport chips through its factories.

Like all chip makers, “they have a lot of visual inspection machines on the factory floor that scan wafers at different stages – and they do a good job of finding anomalies – but they don’t have as well categorized the things they found into types of defects,” said Quinn Killough, Landing’s client liaison.

And like many chipmakers, he had tried a variety of classification software. “But the solutions had to be tailored for each product and with over 100 products, the investment wasn’t worth it,” said Killough, who has a background in computer vision and manufacturing.

AI automates inspection

Then the client applied the AI ​​with landing target. It is designed to manage the end-to-end MLOps process – from data collection to model training and deployment – and then manage the ongoing process of refining models, and especially data, to improve results.

Although it’s still early days for rollout, the product and its data-centric approach have already helped the chipmaker cut costs.

“The lead engineer leading the project said he sees deep learning as transformational and wants to scale it across his entire facility and get other factories to adopt it” , said Killough.

Cloud Inspectors

The chipmaker used LandingLens on NVIDIA V100 GPUs in a cloud-based service that runs inference on hundreds of thousands of frames per day.

“We weren’t sure about the throughput capabilities at first, but now it’s clear it can handle this and more,” Killough said.

The same service can train a new classification model in less than a minute using around 50 defect images so users can iterate quickly.

“On the training side, it’s very important that our tool is fast so that our customers can troubleshoot problems and experiment with solutions,” he said.

Bringing AI to the edge

Now the company is taking the AI ​​work into the factory with a new product, LandingEdge, which is in beta testing with several customers.

It captures images from cameras and then runs inference on industrial PCs equipped with NVIDIA Jetson AGX Xavier Modules. Information from this work feeds directly into controllers that operate robotic arms, conveyor belts and other production systems.

“We aim to improve quality checks, creating a flywheel effect for fast, iterative AI processes,” said Jason Chan, product manager for LandingEdge.

Accelerate the growth of a startup

To gain rapid access to the latest technologies and expertise, Landing AI has joined the NVIDIA Metropolis Programfor businesses using AI vision to make spaces and operations safer and more efficient.

Enterprise and data-centric AI are still in their infancy, which Ng says could be one of the biggest technology shifts of this decade.

To learn more, watch a GTC session (free with registration) where Ng describes the state and prospects of the data-centric AI movement.

Previous

Mandaue factory worker dies while working; the police are investigating the incident

Next

Why This Transparent Factory Was Built In The Middle Of A Forest

Check Also