There’s no doubting that Google Pixel devices have stunning cameras – with the Pixel 2 being the best smartphone camera we’ve ever tested. This is essentially because of the superior image processing, combined with great hardware and powered by Google’s secret AI sauce. Well it’s not going to be a secret much longer as Google has open-sourced part of the code that makes the camera so great.
Google has released a block of code called “DeepLab-v3+” into the public domain as an open source software that will allow manufacturers to bring Pixel-like camera features to their own devices, and help app developers learn how the process works.
As detailed in the blog post by Google, the algorithm uses a neural network to differentiate between the foreground and the background. The image segmentation technology is used to sketch an outline of the objects in the background and the foreground, to create a bokeh effect in the image.
As clarified by Google, this isn’t the exact same technology that Pixel 2 uses – the one the latest Google flagship is more refined and sophisticated to suit the hardware on the smartphone. This algorithm, however, does share the basic features and will require some development effort on the part of OEMs to make it gel with the hardware using their image processing.
Google cautions that the technology does not necessarily assure that camera apps on smartphones can click as beautiful pictures as the Pixel 2; it certainly doesn’t intend to reveal the entire code yet. The algorithm will, however, open up possibilities for better camera performance which, besides the image processing software, also depends on the quality of the sensor and the onboard chipset.
Nonetheless, it is a great step towards improved smartphone photography and we hope to see many more chipsets equipped with AI capabilities in the near future, especially plunging into the budget segment. This will bring about another big change in smartphone photography and allow manufacturers to omit unnecessary gimmicks like slapping nearly every smartphone with a dual camera – even when it cannot perform as intended.