Google's Computer Vision Aids Augmented Reality

Google has announced on their Blog that they’ve made inroads in computer vision, enabling computers to quickly identify images of 50,000 landmarks with 80% accuracy. Google mentions in the post that the effort is aimed at unlocking information from pixels now that they have such a strong handle on unlocking information in text. The ability to catalog and recognize images of the real world plays into the idea of augmented reality and also has implications for building digital city models.

Image recognition adds another means to locate your position in addition to your compass and GPS receiver. The image pattern recognition in its current form just recognizes a landmark among other similar photos, but with enough images and information, it’s feasible that your camera angle and perspective could provide another means to pinpoint your location. If the computer can recognize your location from an image, there stands to reason that you could access information by zooming into a specific detail of that image.

The computer-assisted cataloging of images will help to cluster the massive amount of images on such sites as Picasa and Panoramio. The ability to recognize, and then locate images, could be compiled into a navigable digital reality along the lines of Microsoft’s Photosynth technology.

The key to this announcement in my mind is the ability to make sense of crowd-sourced image capture. The amount of shared pixel-based information is still in its infancy, with rather poor image quality from cell phones, and often long-delayed uploads to the Web.The current unrest in Iran, and the exciting development of citizen-based journalism from cell phone photos and videos points to the possibility for greater real-time interaction with far-flung locations.

In the future, sharing real-time capture from handheld devices that are automatically cataloged and searchable, will help make us all citizens of the planet. And the ability to move forward and backward in time through the images will provide an incredible insight into the dynamic nature of our surroundings.