When we want to make any type of sensitive information public on the Internet, the most common thing is to pixelate a face, email, address or phone number. However, it may not be as effective anymore if we want to protect the privacy of a person in a photo that we share on social networks or publish on a website.
If we have an out of focus photo or we want to hide some type of information within it, a very recurrent tool that we usually use is Photoshop. The quintessential photo editing program allows us to blur an image or part of its content in different ways. However, artificial intelligence is capable of more and more things, and one of them is deciphering what is hidden under a few pixels, managing to recover faces with a detail that leaves you speechless. Let’s see what these tools are capable of and reflect on whether it really helps to pixelate the face in a photo for security and to maintain privacy of the person.
PULSE reconstructs pixelated faces
That artificial intelligence is able to reveal faces that have previously been pixelated is nothing new, but PULSE does it practically perfectly. It is not the first capable of recovering pixelated faces, but it is the one that offers the best results.
Some Duke University researchers in the United States are responsible for the development of this tool. Its operation consists of taking the pixelated image taking it as a reference to generate many faces in high resolution that may resemble each other. Once it has enough, it pixelates them and compares them to the original pixelated image. Once you try all the possible options as a result it shows images with more details, not limited to blurry photos that can hint at a face. Amazing, isn’t it?
Google AI
Based on an initial model, the AI is capable of generating similar images and scale and improve its resolution. Google is also working on this and a research relatively recent development of its AI department shows how new advances offer surprising results.
Google’s machine learning model is capable of taking a low-resolution photo and scaling it up for amazing detail. The method used by Google is known as diffusion model. A generative method that began to be implemented in 2015. The AI is trained to lower the resolution of the photos and take them to extreme pixelation. From this point on, he learns to turn this process around. A method with which the company has managed to improve especially portraits of people. Undoubtedly, some spectacular results that could reveal the identity of any person.
online tools
Even free online tools without any download can expose a pixelated face. Let’s Enhance Thanks to AI, it significantly improves pixelated photos for free. Effortlessly process low-quality images automatically and quickly.
As you can see, once we know what these types of tools that work through AI are capable of, it is quite questionable that it would be of any use to pixelate a person’s face to protect their internet privacy.