Google has issued a clarification claiming the advanced AI-based image processing software it has open sourced isn’t the same that produces the Portrait mode on Pixel 2 handsets. This comes on the back of earlier reports of Google having open sourced the very software that is responsible for the stunning Portrait Mode on the Pixel 2 phones.
That said, Google has however stated the results from the open sourced software can be equally impressive as well. Specifically, it is the whole bunch of code for DeepLab-v3+ that has been open sourced.
Elaborating further, Google said DeepLab-v3+ is essentially an image segmentation tech that helps in creating a distinction between the foreground and background objects in the same image. Google said this is achieved using convolutional neural networks or CNNs so that the outlines of the foreground objects are marked out accurately.
In fact, the more accurate the software is in outlining the target object; more dramatic will be the effect. The Portrait Mode on the Pixel 2 does an excellent job at this and has already gone for rave reviews. With the outline of the objects thus defined, it would be easier to implement that blurring effect on the background while keeping the principal object in sharp focus.
The above has also come to be referred to as the Bokeh effect popularised by Apple. However, while Apple relies on a dual-lens camera for the said effect, that Google manages to do the same using a single lens camera speaks volumes of the efficiency of the software.
Google has said their motive to open source the software is to let industry as well as app developers to use the image segmentation tech for applications they might not have conceived of. As to what practical or real-world uses the software can find application in is anybody’s guess at the moment. Further, too much of a commonality can also ruin the appeal of the background blurring effect as well.