I've just came back after looking at the portrait sample photos published by Google. The Bokeh quality of these photos is awe-inspiring. I just can't believe that a single-camera with Google's machine learning can produce such accurate results.
Now, it's not perfect by any means. I was able to detect some imperfections. For example, look at this sample image of the guy in the street. Now take zoom all in and take a look at the back part of his hair at the top most part. You can see that the blur applied on the hair. The blur also covers the left arm in the edges across it, which doesn't look lie a natural Bokeh compared to an optical one.
That being said, once you zoom out a bit, you can't notice that imperfections and the effect look amazing. Just think about it for a moment, the Google Pixel 2 is able to do that with only a single camera.
Now how does the Google Pixel 2 can achieve that? Well, it has some sort of depth sensing capabilities. It uses a dual-pixel camera sensor. This is the same autofocus technology use in phones like the Huawei Mate 9, Vivo XPlay 6, Moto G5 Plus and many other phones. The sensor has two sub-pixels for each pixels, so each sub-pixel gets to see the image in a slightly different angle. This is used by the Google's machine learning algorithms to understand the scene and produce a depth map.
Just take a look at the image with the girl and the orange hat. Look at the top part of the hat, look how accurate Pixel 2 was able to produce the Bokeh there, it didn't even blurred the strings of the hat there.
The results may vary of course, but this just shows us how amazing is Google'a AI technology. I think the Google Pixel 2 phone is a very important phone for Google. Google has a big advantage over its competitor in the AI department. I knew this for a long time and I didn't know why Google just don't emphasize this and take it to the next level. Now it did it with the Pixel 2 and calling it AI First. This means that Google now goes full force with the AI features.
In the future, those AI features will be the most important once. Even at the present, you can see that the hardware changes aren't really important any more. They don't bring really innovative features any more, but AI will. Google is the leader in the field of Machine learning and we are going to see more and more advanced features taking use of this amazing advanced technology. It will also become better and more accurate. So even this Portrait mode is going to be more accurate as it keep learning from the photos you take and Google will keep perfecting its algorithms.