 |
| Image: online source |
With the help of artificial intelligence, it is able to replace closed eyes with open ones. Facebook Research has developed a system that independently analyzes traditional photographs and is able to automatically replace closed eyes with open ones with the help of the artificial intelligence without the people who see them realizing the change.
Facebook researchers have published a document in which it presents the Exemplar GANs system, based on the Generative Adversarial Network (GAN), an automatic learning technology capable of recognizing faces, while on the other hand, it creates images independently that improve in realism with respect to the previous techniques.
In this case, the AI of Facebook recognizes the shape of the eyes, the color and other features of the physiognomy of each person in the photograph to later replace their closed eyes with open ones.
Facebook's External GAN technology is a variation of the already existing GAN technology that adds additional features and manages to replace environments such as the horizon or a landscape that modify human faces.
On the other hand, Exemplar GANs takes into account the color of the face and the area of the eyes in order to make a more faithful copy of reality, while the GAN system offers results that people recognize as modified images that are imperfect.
In the tests carried out with the new technology, people confused false pictures with open eyes with real ones or did not manage to assure which was the original photo, according to the researchers in the study.
The system still makes mistakes in certain situations, such as when a person's eye is covered with a lock of hair, while at other times the color is not recreated correctly. Facebook Research considers however that these problems are likely to be solved in the future.
Facebook researchers have published a document in which it presents the Exemplar GANs system, based on the Generative Adversarial Network (GAN), an automatic learning technology capable of recognizing faces, while on the other hand, it creates images independently that improve in realism with respect to the previous techniques.
In this case, the AI of Facebook recognizes the shape of the eyes, the color and other features of the physiognomy of each person in the photograph to later replace their closed eyes with open ones.
Facebook's External GAN technology is a variation of the already existing GAN technology that adds additional features and manages to replace environments such as the horizon or a landscape that modify human faces.
On the other hand, Exemplar GANs takes into account the color of the face and the area of the eyes in order to make a more faithful copy of reality, while the GAN system offers results that people recognize as modified images that are imperfect.
In the tests carried out with the new technology, people confused false pictures with open eyes with real ones or did not manage to assure which was the original photo, according to the researchers in the study.
The system still makes mistakes in certain situations, such as when a person's eye is covered with a lock of hair, while at other times the color is not recreated correctly. Facebook Research considers however that these problems are likely to be solved in the future.
No comments:
Post a Comment