Google Starts Using Computer Vision To Search Photos

Google almost completely revamped the Google+ photo experience last week, but somehow the company didn’t get around to announcing one of the coolest photo-related features in its repertoire yet: Google now uses computer vision and machine learning to let you search your own photos for things like sunsets, food and flowers. I also tried terms like “cars,” “beach” and “bikes” and Google consistently returned the right results. This search is built into Google+, but you can also use the regular Google search and use the query term [my photos of xyz] to find the right images.
That’s a huge step forward for photo search in Google. As Google rightly notes, “searching for your photos can be challenging because the information you’re looking for is visual.” I know I’m anything but diligent about tagging my photos, so this new search feature actually allowed me to find random images I had uploaded to Picasa Web a long time ago.
As Google’s Vic Gundotra noted when he announced the new features for Google+ Photos at I/O last week, Google wants to help its users manage their photos. “Organizing photos is often a hassle,” he said, but oddly enough, the company didn’t announce this search feature at I/O and instead waited a week before launching it.