Google is a company who doesn't need any introduction. Its Pixel lineup of smartphones offers the purest Android experience ever. The Google Pixel 2 and Pixel 2 XL might not sport dual camera setup at the back or front. However, the devices comes with the potraight mode at both front and back. Google manages to deliver the same experience driven by AI and software. This has led the camera to be rated one of the best on a smartphone ever. Now the technology has been made open-source by the company.
Google’s research and development team just put out a blog post. The blog post states they are making the open source release of their “semantic image segmentation model”. The model is known as “DeepLab-v3+” and implemented in TensorFlow. Also, according to the blog post, “semantic image segmentation” stands for “assigning a semantic label, such as ‘road’, ‘sky’, ‘person’, ‘dog’, to every pixel in an image.” The company this new application is helping power various other new applications. These new applications include the synthetic shallow depth-of-field effect. This is also known as the portrait mode seen on the Pixel 2 and Pixel 2 XL smartphones.
So how exactly the Google's tech works?
The blog post by Google explains what happens when each pixel or subject in the image is assigned one of these labels. This technology also helps in figuring out the outline of the objects. This is the most crucial thing in the Portrait mode concept. In the Portrait mode, everything apart from the object in Focus is blurred. This gives it a shallow depth of field effect. Apple introduced this in smartphones with the iPhone 7 Plus. While most of the companies rely on a dual camera setup for this, Google did it with just software.
Also Read: How to record the screen on a MacBook without any software
Google’s blog post also says with the DeepLab-v3+ open source release, it includes some more models. These models also include “models built on top of a powerful convolutional neural network (CNN) backbone architecture for the most accurate results…”. The post also states that the models have improved dramatically due to improvement in datasets and more. The post also says this technology will help industry professionals and students will be really benefitted by this.
