AI Gives Visual Anonymity a New Twist
Jeffrey Heimgartner posted on August 07, 2018 |

Artificial intelligence (AI) continues to show that its uses may be never-ending. A research team at Simon Frasier University’s School of Interactive Arts and Technology recently unveiled a new way that AI can be used to modernize how people’s faces are blurred on the news and other television programs.

“Our goal is to create a working technique that would be much better at conveying emotional and knowledge information than current anonymization techniques,” said Steve DiPaola, professor and pioneer in AI/VR facial recognition processes.

Traditional blurring for anonymity requires specialized software that tracks certain areas in the image and adjusts the pixel size and shape based on movement. While this method can effectively hide the face, it can be visually unappealing and neutralize the emotions in a story.

A traditional blurred image, left, verses AI-enhanced anonymity, right. (Image courtesy of Simon Frasier University.)
A traditional blurred image, left, verses AI-enhanced anonymity, right. (Image courtesy of Simon Frasier University.)

Through its research, the team developed a new technique that paints a picture of every frame instead of changing the pixels.

“When artists paint a portrait, they try to convey the subject’s outer and inner resemblance,” DiPaola said. “With our AI, which learns from more than 1,000 years of artistic technique, we have taught the system to lower the outer resemblance and keep as high as possible the subject’s inner resemblance. In other words, what they are conveying and how they are feeling.”

The team’s system uses five levels of AI processing to simulate a smart painter. The AI detects parts of a subject’s face, and then resizes it and the subject’s eyes. After the subject’s appearance is altered, the AI continues to create a picture in several steps, which eventually leads to an image that is reminiscent of works by Vincent van Gough.

To help ensure that the emotions of the situation come through in an image, the AI calculates the facial features and voice tone, and then matches them with the altered appearance. Instead of seeing emotion through a subject’s eyes or mouth, the anonymous face displays it through colors and tones.

“It’s actually an open and dynamic process that allows levels of control throughout,” DiPaola said. “We eventually want the subject or producer to be able to customize the final result based on their needs.”

The team put out a video to give people a better understanding of the process and an opportunity to see it in action.

Interested in learning about more AI innovations? Check out This Deep-Learning AI Can Code Just Like a Programmer and Mystic - AI Controlled Drone.


Recommended For You