Google LLC today introduced new object detection features for AutoML, its cloud-based artificial intelligence development suite, to lure companies working on projects such as augmented reality apps and robots.

AutoML is a set of services aimed at making custom AI software more accessible. It targets companies whose needs are not met by pre-packaged machine learning models, but which don’t have the resources to build a tailor-made algorithm.

The technology provides a middle ground: It allows developers to take a prepackaged Google model and train it for their specific use case using custom data. That brings performance more on par with that of a tailor-made algorithm.

AutoML Vision Edge is a service in the suite that provides the ability to build computer vision software for connected devices. With today’s update, users can train their algorithms to recognize objects in images, a specialized but highly versatile capability.

One area where it can come handy is manufacturing. A factory operator, for instance, could use AutoML Vision Edge to build an AI that spots defects on components traveling along a conveyor belt. A retailer might develop a model that can automatically take stock of the items on store shelves.

There are applications in the consumer market, too. Tryon Technology Inc., an augmented reality startup, is using AutoML to power an app that lets users see what jewelry in a store display would look like on their hand. “By integrating AutoML Vision Edge Object detection into our platform we expect to double our productivity by reducing the amount of resources and time for managing internal infrastructure,” Tryon co-founder Andrii Tsok was quoted by Google as saying.

Another new feature the company unveiled today is object tracking support in its AutoML Video service. The service works similarly to Vision Edge, except it’s optimized for building AI models that process video. Users can now create algorithms that not only identify objects in footage but also track them as they move from frame to frame, which Google sees being useful for applications such as traffic monitoring systems and robotic navigation software.

Image: Google