- It often happens that we take an image but its output is blurred. Now a technology has been developed that will transform a portrait image into a perfect selfie with an Artificial Intelligence tool.
Researchers have designed an artificial intelligence-powered tool that transforms a blurry (blurred) portrait image into a perfect selfie. This new method is named PULSE. This tool matches high-resolution faces generated via AI to an input image that looks similar when compressed to the same size.
According to a report by DailyMail, this team used a machine learning tool with two neural networks. One network develops a human face made of AI, while the other network takes this output and decides that the face looks like the real one.
Through this system, a 16×16 pixel image of a face can be converted to 1024 x 1024 pixels in a few seconds. That is 64 times of a resolution.
The first technology to create a super-resolution image
The computer scientist who led this team of Duke University says, “Never before have such a resolution produced super-resolution images on this resolution.” This system uses multiple pixels to design real-looking faces with imaging features such as fine lines, eyelashes.
Scientists say that this technology is not able to identify people. Researchers say, “It cannot convert a security camera’s out-of-focus, unrecognizable image into a clear picture of a person. Rather this technology generates new faces that are not present but it seems that people with such faces are present. ‘
PULSE received the highest rating
The team used a machine learning tool called ‘generative adversarial network’ (GAN). The first network improves the image through experience until the second network can tell the difference.
Pixelated photos, in which eyes and mouth were barely recognizable. The researcher said that his algorithm could do a lot with this picture which was impossible to do through the traditional approach till now.
Researchers asked 40 people to rate 1440 PULSE-generated photos. Apart from this, 5 other methods were also asked to give a rating between 1 to 5. PULSE received the highest rating.
Thanks For Reading, Credit: Navbharattimes