Google Pixel 2’s portrait mode is now open source

Google recently announced in a blog post that the code for their “semantic image segmentation”, the technology behind the Pixel 2 and Pixel 2 XL’s portrait mode, is now open source.

Semantic image segmentation is the process of assigning a label (e.g. road, sky, person, dog)  to each and every pixel in an image, allowing subject and background to be distinguished from one another. The camera can then blur the background accordingly now that it has been identified.

Google making the semantic image segmentation model open source paves the way for other phone manufacturers to take advantage of it, possibly removing their need to use a secondary camera to map out depth. As we know, the Google Pixel 2 has an amazing portrait mode, and is able to do this with just a single camera lens.

Portrait mode is just one application of the semantic image segmentation model. You can check out Google’s blog post for a more detailed explanation as to how it works.