Is it possible to restore a photograph that has suffered the ravages of time? A new tool uses Artificial Intelligence to achieve unique results.
Currently, the vast majority of photographs that we take are stored in the smartphone or, in the best of cases, hosted in a cloud storage service. This, however, has not always been the case. To go to images of your ancestors, you will most likely have to go to albums. Sometimes, the passage of time has been able to wreak havoc. A bad fold can mean that a photograph is damaged, something that is very difficult to solve due to the mark it has on the paper.
How to enjoy a simple restoration? Artificial intelligence systems, once again, can contribute significantly. There is a tool with which you can enjoy incredible results after a lot of effort made by its creators. This curious technology is called GFP-GAN in honor of its acronym: Generative Facial Prior-Generative Adversarial Network. We are faced with a proposal that manages to carry out a very natural and effective result.
Let us see, therefore, how this curious deep learning based system to achieve incredible results, why it is a tool specially designed to offer maximum performance regardless of the state of the original photo on paper and, of course, to what extent it can be useful to you due to its great usability.
GFP-GAN, the AI-based tool to restore your old photos
The method developed by GFP-GAN is based on a analysis especially focused on the eyes and mouth. From there, he is able to build the rest of the photograph. Thanks to this alternative, it is possible to achieve a natural and, of course, efficient result, capable of eliminating the imperfections present in physical photography. According to the information available on Engadget, this technology aims to create a pattern to later digitally reflect all the imperfections present in the physical support. Any bends are replaced locally to reflect an image that would copy the original model.
It is a technology that, unlike the rest of the alternatives on the market, has a series of additional tools to work in depth the aspects that define its level of realism. It is important to point out that, however, the problems related to quality are not solved if you decide to enlarge the size of the image that you want to print. If this happens, a series of imperfections will be seen that, according to the project that is being developed, it is expected that they will be solved in future updates.
At the moment, this technology is available for free. Similarly, the code for this tool is publicly available, so Significant progress could be made in the coming months. In fact, it is expected that successive changes will arrive in order to improve the quality of this proposal and, above all, the improvement of the sharpness related to the change in size of the digital model.
Related topics: Technology