বৃহস্পতিবার, ১৬ ফেব্রুয়ারী, ২০১২

A Sneak Peek Inside Four Silicon Valley Tech Labs

Project: Improving the Google Goggles Recognition Engine

Several years ago, Google engineers started wondering how your phone could provide real-time data such as identifying a building, tree, or person near you. Their first effort, called Google Goggles, could read a book cover and offer a description and price. Point your phone at a postcard and the app could identify the Statue of Liberty and offer some facts about the landmark.

Unfortunately, the app was limited to 2D objects with an obvious shape or easily readable text. Now, researchers at the search giant have decided to expand the app's capabilities. Their most recent enhancement: Google Goggles can identify foreign currency, such as a euro or a peso, and look up the exchange rate between that currency and the dollar. The app also has a continuous scanning mode so you don't need to snap a photo of the object.

Hartwig Adam, whom I met at the Google headquarters in Mountain View, Calif., is the tech lead for Google Goggles. He says his team has high ambitions for the app's future: he envisions the app being able to scan trees and plants and tell you what species you're looking at, or look over at a passing car and tell you the make and model.

It's a little like a visual version of the popular app Shazam, which many iPhone users deploy to name songs playing in the background. But visual recognition is more complex. "What humans do effortlessly, such as recognize a car or a cat, is hard for machines to do," Adam says. While a song-identifying app has to filter out noise, the song it's recognizing sounds the same every time. But each red oak tree looks a little different and it looks a little different from second to second, as the breeze blows through its branches. To pick out a car for you, Google Goggles would not only need to know the difference between a Toyota Camry and a VW Passat but also recognize the Camry from whatever point-of-view you are observing it.

The app uses algorithms that determine the size and shape of objects. It sends a low-res version of scanned images to a Google server for processing. Then the app shows the name of the object and a description. If anything helps Google Goggles to improve further, Adam says, it might be the introduction of very high-resolution cameras?such as 16-megapixel?into smartphones, which would give the app hi-res images to work with.

Google has some competition in the space, however. A relatively obscure app called Word Lens can read text on a sign and convert it to a different language. Another app, oMoby, also scans objects and returns a text description but that app seems to have languished. But, as usual, Google's main competition seems to be itself: While the company did not mention it to us directly, Google is supposedly working on augmented reality glasses that show you information about your surroundings.

Source: http://www.popularmechanics.com/technology/gadgets/news/a-sneak-peek-inside-four-silicon-valley-tech-labs?src=rss

donald driver donald driver martin henderson mlk day octavia spencer golden globes 2012 winners golden globes 2012 red carpet

কোন মন্তব্য নেই:

একটি মন্তব্য পোস্ট করুন