What Does A.I. Have To Do With This Selfie?

preview_player
Показать описание

Recently, you may have noticed people sharing stylized photos and videos that resemble famous paintings, like "Starry Night" by Van Gogh and "The Scream" by Munch. But how do the apps that make these images work? How can the style of a painting be transferred to a selfie or a photo of your dog? Turns out, a whole lot of A.I. using machine learning and deep neural networks. In this video, we break down the basics of how style transfer works, and demo some recent style transfer experiments created by research scientists at Google.

Style Transfer research papers referenced:

In addition to internal Google experiments and tools, some style transfer images and videos created using these apps:

Thanks to Vibe Mountain for the excellent musical styles in this video. And thanks to Beardy Glasses & MixTape Club for the excellent illustration and animation styles.

-------

:)
Nat & Lo
Рекомендации по теме
Комментарии
Автор

At 3:38, it is mentioned "we take a random white noise image and keep adjusting the pixels until we can pass into the same neural net .. ", how do we do that?

nitinsurya
Автор

I wonder how DNN can be applied to audio & speech. I imagine being able to train it on the speech patterns of a particular person, learning their speaking style, and then I speak into it (or type) and it comes out sounding like that person. Sadly it has many evil applications, but I imagine adding this to what they did with Audrey Hepburn on the bus and you can bring dead movie stars back to life.

jtgdyt
Автор

what exactly are the inputs and outputs of training such a transfer style network?

DivineKage
Автор

Nice video. Have more knowledge about style transfer

liccoweeee
Автор

Like always, I love the content and how you explain complex information in a very simple fashion. My only disappointment is your using an iPhone instead of a baller new Pixel phone. Google hook her up.

MarkBurress
Автор

I'm a little leery about the explanation stating that artificial neural network (ANN) layers look at various parts of an image, since as I understand that's not the way it works. I suppose it's not a bad way to explain it, but it's just not true (from what little I understand about the subject). Please, anyone, correct me if I'm wrong and please show proof.

Soulsphere
Автор

I'm guessing Lo wasn'tt too interested in this :p

sshum
Автор

i gues that is what apple dit to its last iPhone they used an image from their previous

Chrissleijpen
Автор

Why is it that all sentences begin with "So" these days?

JustAnotherPersonHere
join shbcf.ru