Amazon's Smartphone Event | CNET

Amazon's Smartphone Event

Join CNET for coverage from Amazon's event, starting at 10:30 a.m. PT on Wednesday, where the online retailer is widely expected to unveil its first smartphone. Our live blog will kick off about an hour beforehand and will bring you news updates, photos, and commentary.

  • Roger Cheng 6/18/2014 6:34:26 PM
    You can tilt the phone to look at the different perspective in the game.
  • Jessica Dolcourt 6/18/2014 6:34:39 PM
    The block of tofu also looks like a tasty gourmet marshmallow.
  • Roger Cheng 6/18/2014 6:34:47 PM
    This game looks trippy.
  • James Martin 6/18/2014 6:34:48 PM
  • James Martin 6/18/2014 6:35:03 PM
  • Jessica Dolcourt 6/18/2014 6:35:05 PM
    I just had fried tofu last night. It was delicious.
  • Roger Cheng 6/18/2014 6:35:17 PM
    It's pretty impressive, you just move your phone around and you move the perspective around.
  • Roger Cheng 6/18/2014 6:35:25 PM
    It does seem like it would be pretty natural.
  • Roger Cheng 6/18/2014 6:35:49 PM
    When we show this to people for the first time, the first question they ask: how do you even do that? he says.
  • James Martin 6/18/2014 6:35:52 PM
  • Roger Cheng 6/18/2014 6:35:54 PM
    The key is knowing where the head is at all times.
  • James Martin 6/18/2014 6:36:06 PM
  • Roger Cheng 6/18/2014 6:36:20 PM
    We started working on this four years ago. We had early prototypes working in the first week, he says.
  • Jessica Dolcourt 6/18/2014 6:36:21 PM
    This is the kind of eyeball tracking that Samsung wished it had done instead when it released the Galaxy S4.
  • Roger Cheng 6/18/2014 6:36:46 PM
    Early prototypes required users to wear headgear that allowed the phone to track where the head was positioned.
  • James Martin 6/18/2014 6:36:48 PM
  • Roger Cheng 6/18/2014 6:37:00 PM
    The right way to do this is with computer vision, he says. It's a tough problem.
  • James Martin 6/18/2014 6:37:06 PM
  • Roger Cheng 6/18/2014 6:37:19 PM
    In the real world, you have different factors like lighting, shadows, beards, glasses, make-up, etc.
  • James Martin 6/18/2014 6:37:20 PM
  • Jessica Dolcourt 6/18/2014 6:37:23 PM
    Amazon has been working on this for 4 years, but long before that, this kind of eye-tracking research has lived in the realm of university labs at the likes of MIT.
  • Roger Cheng 6/18/2014 6:37:29 PM
    For a commercial product, it has to be robust, he says.
  • James Martin 6/18/2014 6:37:51 PM
  • Roger Cheng 6/18/2014 6:37:59 PM
    Using the existing front-facing camera won't work. The field of view is too narrow, he says.
  • James Martin 6/18/2014 6:38:05 PM
  • James Martin 6/18/2014 6:38:18 PM
  • Roger Cheng 6/18/2014 6:38:19 PM
    Amazon made special cameras with a much wider field of view, he says.
  • Jessica Dolcourt 6/18/2014 6:38:19 PM
    MIT found that it was really hard to track eyeballs on a Web page, but on a handheld device like a cell phone -- and one that includes its own camera with a wide field of view -- you suddenly have a solvable problem.
  • James Martin 6/18/2014 6:38:46 PM
  • Roger Cheng 6/18/2014 6:38:49 PM
    The phone also needs to know how far away the head is, so Amazon added a second camera.
  • James Martin 6/18/2014 6:38:57 PM
  • Jessica Dolcourt 6/18/2014 6:39:17 PM
    The two cameras are called stereovision, BTW. But it doesn't stop there.
  • James Martin 6/18/2014 6:39:26 PM
  • Roger Cheng 6/18/2014 6:39:28 PM
    The problem is users tend to hold the phone in different ways, which may obscure one of the camera. So it added four corner cameras.
  • Roger Cheng 6/18/2014 6:39:40 PM
    But then people use their phone in absolute darkness, whether up at night or in car.
  • James Martin 6/18/2014 6:39:45 PM
  • Roger Cheng 6/18/2014 6:39:57 PM
    These phones have to work in the dark. So it added infrared lights, one in each camera.
  • Jessica Dolcourt 6/18/2014 6:40:16 PM
    At any given time, at least two of the Fire Phone's four corner cameras and center camera can find your eyes, creating stereovision.
  • James Martin 6/18/2014 6:40:20 PM
  • Roger Cheng 6/18/2014 6:40:20 PM
    They are global shutter cameras instead of rolling shutter, which means they are 10 times more efficient, and are turned on only when we needed.
  • James Martin 6/18/2014 6:40:36 PM
  • James Martin 6/18/2014 6:40:54 PM
  • James Martin 6/18/2014 6:41:25 PM
  • Roger Cheng 6/18/2014 6:41:26 PM
    There are other wrinkles, including baldness, sunglasses, etc, other varieties that make this a tough machine learning problem, he says (his fav problem is baldness, of course).
  • Jessica Dolcourt 6/18/2014 6:41:37 PM
    So, to recap. There are four corner cameras an a center camera, each with an infrared light to combat the dark, and also a huge image database to "train" the tool's algorithms.
Powered by ScribbleLive Content Marketing Software Platform
This week on CNET News See all