Facebook

This is Pulse, the tool that reconstructs pixelated faces

Pinterest LinkedIn Tumblr

Pulse is an Artificial Intelligence (AI) image processing tool capable of recovering blurred or pixelated faces in great detail. The tool has been developed by a group of researchers from Duke University in the United States.

With this type of tools it is possible to recover images that have been pixelated to protect the identity of a person

Pulse is not the first or the only tool of this type, which bases its operation on an AI, however it has two distinctive features. The first of them is that it is able to reveal faces almost perfectly despite being pixelated. This is due to the second peculiarity of the tool, which is the way in which it carries out this process.

The process Pulse uses to retrieve faces is different. First, it takes the pixelated image and uses it as a reference to generate many high-resolution faces that resemble the pixelated original. Once it has generated a sufficient number of faces, it pixelates and compares them with the original, testing all the options it has generated. The result is more detailed and realistic images. In the following image you can see the process of generating faces and comparing them with the original pixelated:

Generally, these AI-based tools when they recover a pixelated image, what they do is analyze the pixels that surround an area and generate similar ones. This process has a drawback and is that they often fail to guess what exactly the pixel it analyzes is like and what they do is imitate other pixels, creating blurred or flat images . This is why many deepfakes are strange and unnatural.

In the following image you can see how Pulse works (the last two columns) compared to other similar AI imaging tools:

The application of Artificial Intelligence to image treatment and recognition is an interesting field, with applications such as Google Lens Augmented Reality capable of converting handwriting into a text file or bringing a zoo to the user's home. AI can also be applied to the random generation of images of people or even cats, although in these cases the result leaves some nightmare images.

Tom Walsh

Tom Walsh is a social media marketing professional with 30+ years of experience. Previously, Tom worked as a Facebook marketing manager for a giant tech startup. He graduated with honors from Columbia University with a dual degree in Business Administration and Creative Writing."