Google Pixel 3 implements Artificial Intelligence within Portrait Mode
This is how Artificial Intelligence is implemented in the new Portrait Mode that we can find in the Google Pixel 3 that have just been launched on the market.
Artificial Intelligence is creeping into all kinds of products due to the great technical and technological advantages that it can offer us. One of the fields where this technology is being exploited the most is in photography and in this case Google has published on its blog how the AI inside the Google Pixel 3 has the ability to measure the depth of the Portrait Mode that are accompanied by algorithms of Machine Learning.
Google explains how Machine Learning improves photos on the Pixel 3 smartphone.
The Pixel 3 has Portrait Mode that has a function that blurs the background depending on the depth that allows attention to be focused on the person or object in the foreground.
Using Google's AI called TensorFlow and which is open source, the company's engineers can train a convolutional neural network that collects information from the different pixels of the PDAF approach and then has the ability to predict depth. . Through this renewed and improved depth estimation method that is based on Machine Learning, it offers better quality in Portrait Mode of the new Pixel 3.
“To train the neural network, you need a lot of images from the PDAF approach and the corresponding high-quality depth maps. And since we want our predicted depth to be useful for Portrait Mode, we also need the training data to be similar to the images users take with their smartphones.
To achieve this, we created our own custom platform, the “Frankenphone”, containing five Pixel 3 phones, along with a WiFi-based solution that allowed us to capture images from all phones simultaneously (within a tolerance of ~ 2 milliseconds ). With this platform, we calculate the high-quality depth of the photos by using a multi-view stereo and movement structure, ”the company explains on its official blog.
Source: HOCP
