Baidu Is Bringing Intelligent AR to the Masses
An augmented-reality system powered by computer vision and deep learning will add an extra layer to the real world for millions of people.
The Chinese Internet giant Baidu has announced a new AR platform that will allow people to make use of the technology from within many of the company’s own apps.
The system, called DuSee, is claimed to make use of "sophisticated computer vision and deep learning" to “understand the 3-D environment, and create virtual objects that have rich interactions with the user and the real world.” The platform seems to sit somewhere between basic AR systems, like those used in Pokémon Go, and more advanced technologies, such as Google’s Tango. The former simply drops animations onto a device’s screen to make it look like they’re in situ without any analysis of the real world; the latter uses special 3-D camera technology to map the local environment and overlay visual graphics.
Instead, DuSee seems to make use of a smartphone’s single camera, using computer vision software to detect and interpret what it sees in order to add a layer of extra information for the user.
Baidu has offered up two examples of the system in use. One shows the software recognizing the presence of a 2-D map of Shanghai, then render it as a 3-D illustration reminiscent of SimCity. Another identifies a shampoo advert, adding virtual petals that echo its floral branding while keeping them locked in place as the ad moves in space.
Perhaps most exciting, though, is the fact that Baidu plans to bake this platform into its most successful apps, including its Mobile Baidu search software. That app alone has hundreds of millions of users, so the AR will find its way into the lives of many people, very quickly.
Baidu has been doubling down on its development of AI systems in recent years. While in the past it’s flirted with some goofy-yet-impressive image manipulation software, it’s also been using its homegrown Paddle deep-learning system to create more useful products and prototypes. Examples include a headset that uses computer vision to convert images to audio that tells visually impaired people what is in front of them, and powerful speech recognition for its voice interfaces. Indeed, the company points out that it’s “excited about future directions such as integrating AR with speech recognition and natural language processing" into DuSee.
With its cute characters and compelling gameplay, Pokémon Go—which hasn't yet launched in China—has whetted the world’s appetite for AR. Baidu, though, may be the first to offer a truly intelligent augmented reality experience to a mass audience.
Keep up with the latest in AI at EmTech Digital.
The Countdown has begun.
March 25-26, 2019
San Francisco, CA