Home News Google Shares Technology Behind Portrait Light, Enhances Portrait Lighting With Machine Learning

Google Shares Technology Behind Portrait Light, Enhances Portrait Lighting With Machine Learning

To allow people to better emulate professional-looking portraits, Google recently released Portrait Light for the Pixel Camera and Google Photos apps for smartphones. The new feature enhances the illumination for the captured portraits by adding a simulated directional light source, with the directionality and intensity set to complement the lighting from the original photograph. Traditionally, professional portrait photographers use specialized equipment like off-camera flashes and reflectors to create such compelling photographs.

Google Shares Technology Behind Portrait Light, Enhances Portrait Lighting With Machine Learning
GIF courtsey: Google AI Blog

Pixel Camera on Pixel 4, 4a / 4a 5G, and Pixel 5 automatically applies Portrait Light to images post-capture in the default mode. Additionally, the feature also applies to Night Sight photos that include people (one person or a small group) automatically. It provides more dynamic lighting to Portrait Mode photographs while accompanying the already-applied shallow depth-of-field effect, thereby creating studio-quality portraits. On the other hand, in Google Photos, users can manually re-position and adjust the brightness of applied lighting.

Technology Behind Portrait Lighting

Google Shares Technology Behind Portrait Light, Enhances Portrait Lighting With Machine Learning
GIF courtsey: Google AI Blog

Google has presented the technology behind Portrait Light that uses machine learning models to imitate off-camera lights used by portrait photographers. “Portrait Light models a repositionable light source that can be added into the scene, with the initial lighting direction and intensity automatically selected to complement the existing lighting in the photo, said Google in a blog post.

The tech giant constructed a custom lighting rig containing 331 individually-programmable LED light sources and 64 different cameras with their own viewpoints. It then photographed seventy different people using the Light Stage computational illumination system to generate the training data. After enough training on these images, Google’s AI can now alter the lighting to appear from a myriad of different directions based on the head position of the portrait’s subject. The algorithms estimate the head pose using MediaPipe Face Mesh.

Google Shares Technology Behind Portrait Light, Enhances Portrait Lighting With Machine Learning
Image courtesy: Google AI Blog

You can try Portrait Light on your existing images with slightly underexposed faces in your Google Photos library. The feature especially benefits the images with a single individual posed directly at the camera. According to Google, Portrait Light is the company’s first step towards creative post-capture lighting controls for smartphone cameras using machine learning.

LEAVE A REPLY

Please enter your comment!
Please enter your name here