Nvidia taught artificial intelligence to cross people’s photos

Three developers from Nvidia have taught the neural network to merge several photos of people into one, keeping common details from each snapshot. The program crosses the features of the face, the shape and color of the hair, the background and even minor details, such as glasses. You can combine two, three and more photos.

From each snapshot, the neural network can select one specific to do and add to the combined snapshot, according to TechCrunch. For example, from the first photo you can take the background, from the second face, and from the third hair color. The result is too realistic:


In detail you can see that the result is not perfect. Although such a photo can be easily used for a profile in social networks, and hardly any of your friends will replace the fact that the picture is not real.

Notify of
Inline Feedbacks
View all comments
Would love your thoughts, please comment.x