Google’s AI-Powered Tool For Excellent Portrait Mode Images On Pixel 2 Devices


Software and search heavyweight Google has open-sourced an artificial intelligence (AI) tool which controls for the remarkable portrait mode in its Pixel 2 devices. Google’s flagship from last year, the Pixel 2 and the Pixel 2 XL pack an impressive camera.

What’s with the AI tool for portrait images

Google has publicized plans to open source the image segmentation tools behind its AI-powered camera Portrait Mode. This means other developers should be able to integrate and utilize the Portrait Mode feature that Google’s Pixel 2 smartphones use. By this means allowing them to create camera apps which provide excellent Portrait shots, though they may be featuring a single camera.

Certainly, Pixel 2 has one of the first quality cameras ever seen in the history of smartphones. Most of what makes the Pixel cameras exclusive has to do with AI as well as machine learning tools crammed inside by Google, thus resulting in high-pitched, more comprehensive images and superb bokeh photos. The AI tool which helps Pixel in capturing great bokehs through one single camera lens has now become an open source.

Need for DeepLab-v3+ models

This AI-powered bokeh code, referred to as DeepLab-v3+, is open sourced and instigated in Tensorflow. This allows the makers of the phone as well as app developers to apply the code to their products. Google software engineers Liang-Chieh Chen and Yukun Zhu stated that DeepLab-v3+ models are built on top of a robust Convolutional Neural Network (CNN) architecture to make available “the most accurate results, projected for server-side deployment.”

Google hopes that public sharing of the system with the community will help in making it easy for the other group’s academia and industry to replicate and additionally improve upon state-of-art systems, train models on the new datasets, and visualize new applications for the technology.

Five months down the line after launching of the device, Google has stated on its Google Research blog its plan to open-sourcing the ‘DeepLab-v3+’, which is its “newest and the best performing model for semantic image segmentation.

On how to make things easier for everyone, this tool will be used in identifying and distinguishing people as well as any other object from the background for portrait mode-like results that allow a blur to be applied to the background layer.


Please enter your comment!
Please enter your name here