Google Opens Pixel's Depth-Of-Field Effect Code to Researchers
Mar 15, 2018, 8:21 AM by Eric M. Zeman
updated Mar 15, 2018, 8:49 AM
Google this week made its semantic image segmentation feature, which is what powers the portrait effect on the Pixel smartphones, available to academia and researchers. Google uses semantic image segmentation to define the shape of objects within photos and then assign those objects a label, such as person, dog, or car. In short, it's what allows smartphones such as the Pixel to recognize the shape of a subject's head in a portrait, draw a line around it, and then take action with the other regions of the photo, such as blur the background. Google is open-sourcing the newest version of its semantic image segmentation code, DeepLab-v3+, as implemented in TensorFlow. It is built on top of a convolutional neural network (CNN) backbone architecture, which Google says is meant for server-side deployments. Google's immediate goal is not to bring Pixel-level portrait shooting to other smartphones. "We hope that publicly sharing our system with the community will make it easier for other groups in academia and industry to reproduce and further improve upon state-of-art systems, train models on new datasets, and envision new applications for this technology," said the company. Researchers and academics interested in the technology can find the necessary resources on GitHub.
Hands On with Android P
Smartphone Camera Shoot-Out: iPhone X, Pixel 2 XL, Galaxy S9+
Review: Tech 21 Evo Check Case for Pixel 2 XL
Holiday Gift Guide 2017
Review: UAG, Speck, and Incipio Cases for the Google Pixel 2 XL
Google Pixel 2 XL
Comments
No messages
