Researchers at Intel have developed an eye contact correction system for video chats.
The essential advantage of video chats over other forms of electronic communication is the, well, video. It means you get to see the other participants and catch all their reactions.
There’s just one small issue.
Maintaining direct eye contact with the other participant during video chat can be a bit tricky. Instead of looking at the screen, users often have to look at the front-facing camera.
Although millions of people are using this communication technology daily, researchers have failed to solve this pesky problem – until now.
A team of researchers at Intel recently developed an eye contact correction model to address this issue. How does it work, you ask?
It basically restores eye contact during live video chats, regardless of the camera placement. Using this new model, users can center their gaze without specifying or redirecting the display angle or user geometry.
In a statement to TechXplore, one of the researchers, Leo Isikdogan said:
“It is hard to maintain eye contact during a because it is not natural to look into the camera during a call. People look at the other person’s image on their display, or sometimes they even look at their preview image, but not into the camera. With this new eye contact correction feature, users will be able to have a natural face-to-face conversation.”
How the Eye Contact Correction System Works
Eye contact correction in video calls is not a new idea.
In the past, researchers have proposed image manipulation models to address the problem. While some suggested setting up specialized hardware, others have used some form of expensive computational processes.
For their new system, the Intel researchers used a deep convolutional neural network (CNN) to warp and tune the eyes in its input frame. That way, they could successfully redirect a person’s gaze.
First, the CNN processes a monocular image. Then, it creates a vector field and a brightness map to correct a user’s gaze.
Unlike previous eye contact correction approaches, the new model can run out of the box without any user input. Furthermore, it does not require any dedicated hardware.
Isikdogan noted:
“Our eye contact corrector uses a set of control mechanisms that prevent abrupt changes and ensure that the eye contact corrector avoids doing any unnatural correction that would otherwise be creepy. For example, the correction is smoothly disabled when the user blinks or looks somewhere far away.”
Isikdogan and his team generated photorealistic and labeled images and conducted bi-directional training for the model. Using a series of blind tests, the researchers then evaluated the model’s effectiveness.
As you can imagine, the test was a success. Now, the researchers are trying to perfect the technology and apply it in existing video conferencing services.
Who knows? This new eye contact correction system could potentially be on your favorite video chat app soon.
Comments (0)
Most Recent