Google incorporates different upgrades to both software and hardware on its Pixel phones, just like the rest of the competition, but Google's case is different. Mountain View group take time to explain to us what you are doing, and why. He did it in Astrophotography mode, in Dual camera Portrait mode and with AI after audio recording, among other things.
Now we just go to a new definition about 4 pixel infrared face unlock. A program that uses dedicated hardware in addition to software, similar to a program installed at the time by Apple with its own ID, but with a & # 39; Google-like & # 39; s twist, and an explanation of the reasons for choosing this program.
Depth maps with parallax
Google explains that the face recognition system was originally proposed by RGB cameras, but that processing these images to serve their purposes would have been "too expensive". In other words, it will require a lot of effort around the phone processor to get a simple result with various applications, which is why- infrared system.
So, Google installed an infrared system based on two front sensors on the Pixel 4, one that will work well and quickly in the dark. He was born The iDepth program, which not only allowed face recognition but also protected identity theft since it was a three-dimensional learning system mixed with various algorithms provided by artificial intelligence.
This program, in fact, has just been opened in other applications thanks to the Camera2 API, so that its readings can be used to enhance previous blurring of selfies, and other opt-in functions. In a very brief way, Google explains that its program is iDepth is based on what is known as parallax. A parallax is nothing more than a direct deviation from the physical property of an object, depending on the point of view.
One light source, two "eyes."
With this idea in mind, Google is coming together one black emitter to put on the front cover of the phone This creates a point system similar to that developed by the Time of Flight system, the TOF sensors we have often described and used for recognition and photography. The difference is that alongside this single output is available two cameras for capturing a bad beam, one on the front side of the phone.
Therefore, the infrared map is read from two points, which helps to use mathematical algorithms to create a three-dimensional wide-angle map. In the same way that people (and many other types) learn the depth of scenes due to the shape of each eye. As Google explains, when a program like Depth is properly calculated, "the reconstruction is generated by metric, which means they express actual physical distances"
An example of a three-dimensional image was created from a single read by iDepth thanks to its dual camera for deep learning and deep learning algorithms and artificial intelligence algorithms.
Google also explains that its iDepth program imagines itself in the event of a slight decrease in logical deviation when scanning optical images in front of a computer. The system is also connected to to study deep or to study deeply responsible for developing raw data. In this way, it is possible to create a very sophisticated three-dimensional map with great speed, and it works just as well with a good light as its exterior.
Google concludes by explaining that iDepth produces a smooth 30Hz real-time map that is compatible with photography and video. This is allows processing to be used later, not only during photo shoot or face recognition during opening hours. The demo app to monitor iDepth performance is on this link, though it only works on Google Pixel 4 as it depends on the phone hardware.
More details | Google
