Today Google has announced the release of MobileNets, a series of TensorFlow vision models built for comparatively low-power, low-speed platforms like mobile devices. In a cross-post on both the Open Source and Research blogs, Google released details about the new visual recognition software. Now even more useful machine learning tools can operate natively on your phone’s hardware, in a fast and accurate way. And, future tools like Google Lens will be able to perform more functions locally, without as much need for mobile data, and without waiting.
Google I/O is underway and one of the first topics covered was the great strides the company has made in image recognition software. Further improvements in computer vision are enabling Google products to see the attributes behind the picture and break them down into constituent parts. This is powered by machine learning and capable of some incredible image editing feats.
Google has shown a lot of interest in computer vision over the last few years. You need only look at the way it digs around for similar images in Google search and identifies specific objects in your uploaded photos. It can always be better, though, so Google has purchased a French company called Moodstocks that specializes in image recognition.
Google's Goggles is all but abandoned now. We've seen Google resurrect apps from the dead and update them after years of neglect, but it's hard to imagine the company putting a fresh coat of paint on Goggles at this point. If only because the app has been superseded by others from Google, with its functionality cut off into little pieces and moved to various places inside the ecosystem.
But that doesn't take away from the fascination and respect that Goggles deserves. It could recognize landmarks before Google Photos, read and translate text before Google Translate, use OCR on images before Now on Tap, and even solve sudoku puzzles, scan and add contacts from a business card, and find and suggest similar products — all options that have yet to be transplanted into any other Google app.
Typing to find things while shopping? What is this, the stone age? Pounce is a new app that claims to be able to identify a product from a photo in mere moments. Then it helps you find similar things to buy right from the app. Does it work? Yeah, mostly.
When it comes to looking for love, finding a match isn't always easy. This is true of people, and it applies to pets as well. Superfish's PetMatch app tries to cut down on the amount of time it takes to find a companion. Users open the app, snap a picture of the kind of dog or cat they're looking for, and peruse a list of similar ones just waiting for someone to come along and adopt them.
Superfish specializes in the development of image recognition software, and the company has come out with a couple of ways of utilizing this capability. WindowShopper (not available for Android) helps consumers find products similar to something they have a photo of, and PetMatch extends the concept to pets.