This neural network knows how you will look in 50 years. How it works?

It is no secret that now the neural network most commonly used in photography. We have already seen how they are able to work with animation on a standard PC, and the picture, written in artificial intelligence, going under the hammer more works of famous artists. It is not surprising that neural networks were the basis of the FaceApp application that in a few days gained popularity around the world, storming the App Store and Google Play.

Elon Musk once colonized Mars

Why photography? The answer is really simple images and a variety of images online a lot. And not just a lot, billions of them! Probably the number of photos exceeded the population of the Earth, and let not all the pictures uploaded to the Internet, they are still very much. Therefore, the neural network much easier to train on the basis of photographs: she can “feed” the multitude of data free. That is how the developers FaceApp managed to achieve impressive accuracy in the processing of the video application is very well able to artificially “sostarivayut” and to rejuvenate people, to change her hairstyle, apply makeup possibilities there quite a lot.

The development of FaceApp is our compatriot Yaroslav Goncharov, a native of “Yandex”, who founded the company Wireless Lab. Several projects of the new company were associated with deep learning neural networks, and “shot” it FaceApp, because photos and videos will be the most important type of information in the near future. Through the analysis of regularities in the appearance of people FaceApp creates a new image, which one looks more young adults or, for example, smiling.

Jeff Bezos once sold by Amazon

The app became popular after a few months after the release in early 2017. FaceApp kept in the top of the App Store and Google Play in Russia, USA, UK, France, Germany and many other countries. Soon, the app reached 40 million downloads. Interestingly, a new round of popularity FaceApp is filter of “aging”. Many young people have been wondering how they will look in 30, 40 and even 60 years, and FaceApp at the expense of accuracy of their algorithms has allowed them to do it.

The developers have not disclosed the algorithm used (unsurprisingly). It is possible that the project is based on the algorithm of generative neural networks StyleGAN developed by Nvidia. He was involved in the project thispersondoesnotexist, which generated realistic individuals who in fact never existed. However, in the event there was the work of two neural networks: one of them was responsible directly for the creation of the image, and the second analyzed the image on the subject of realism. When generating a faces the neural network uses the features of several people, and then adapts them to the desired age, facial expressions, hair length, and other parameters.

In the case of FaceApp is probably the generator first determines the attributes of a human face (the smile to the hair and even freckles), and then imposes on him the attributes of an old/young face (like wrinkles or, on the contrary, smooth skin) but so that the image looked like this.

26-year-old author after decided to meet an old age in Switzerland

Read also — the neural Network of Yandex, wrote a piece for Symphony orchestra

How to download FaceApp

The app is free to download for iPhone (here) and Android (here). Initially available an initial set of tools, but if you subscribe, you will receive advanced filters.

FaceApp collects user data?

When you use the services, which are based on neural networks, need to understand that every “fed” them with megabytes of data is used to further train the neural network. So in a sense, the app actually collects user data. And there are two sides to the coin: with one of your photos I want to keep only for themselves; on the other, the neural network need to be trained, because without that FaceApp could not be so realistic to change pictures.

Of course, good developers must warn that collect the data, otherwise someone might think that subsequently resell them and use the marketers for advertising. Or a photo of the author will appear in the advertisement of one Bank for pensioners (I hope not).

I wonder how far advanced neural networks in the next few years. Just two years ago, the Lyrebird from Montreal have created a speech synthesizer based on the AI, able to replicate any voice. To follow the voice system takes only a few seconds of audio recording of voice required of a person on the basis of which will create an audio snippet. Accurate imitation of the voice is possible through the use of neural networks based on artificial intelligence and operating according to the same principles as the neural networks of the human brain. The AI learns to recognize features of human speech, and then these data are already used for the synthesis of an artificial voice. So selfie is not the limit of the use of neural networks.

But data collection should think.

Leave a Reply

Your email address will not be published. Required fields are marked *