# This New AI is Photoshop For Your Hair! 🧔

## Метаданные

- **Канал:** Two Minute Papers
- **YouTube:** https://www.youtube.com/watch?v=YNY_ZEuDncM
- **Дата:** 01.05.2022
- **Длительность:** 5:12
- **Просмотры:** 88,694
- **Источник:** https://ekstraktznaniy.ru/video/13581

## Описание

❤️ Check out Weights & Biases and sign up for a free demo here: https://wandb.com/papers 
❤️ Their mentioned post is available here (thank you Soumik Rakshit!): http://wandb.me/barbershop

📝 The paper "Barbershop: GAN-based Image Compositing using Segmentation Masks" is available here:
https://zpdesu.github.io/Barbershop/
https://github.com/ZPdesu/Barbershop

❤️ Watch these videos in early access on our Patreon page or join us here on YouTube: 
- https://www.patreon.com/TwoMinutePapers
- https://www.youtube.com/channel/UCbfYPyITQ-7l4upoX8nvctg/join

🙏 We would like to thank our generous Patreon supporters who make Two Minute Papers possible:
Aleksandr Mashrabov, Alex Balfanz, Alex Haro, Andrew Melnychuk, Angelos Evripiotis, Benji Rabhan, Bryan Learn, B Shang, Christian Ahlin, Eric Martel, Gordon Child, Ivo Galic, Jace O'Brien, Jack Lukic, Javier Bustamante, John Le, Jonas, Jonathan, Kenneth Davis, Klaus Busse, Lorin Atzberger, Lukas Biewald, Matthew Allen Fisher, Michael Albrecht, Mich

## Транскрипт

### Segment 1 (00:00 - 05:00) []

Dear Fellow Scholars, this is Two Minute  Papers with Dr. Károly Zsolnai-Fehér. AI research is improving at such a rapid pace,  that today, we can not only create virtual humans,   but, even exert some artistic control  over how these virtual humans should look.    But today, we can do something even cooler.   What can be even cooler than that? Well,   have you ever seen someone with a really cool  haircut and wondered what you would like with   it? Would it be ridiculous? Or would you be able  to pull it off? Well, hold on to your papers,   because you don’t have to wonder anymore. With  this new AI, you can visualize that yourself! Hmm, okay, but wait a second. Previous  methods were already able to do that,   so, what is new here? Well, yes! But look at  the quality of the results. In some cases, it   is quite easy to find out that these are synthetic  images, while for other cases, so many of the fine   details of the face get lost in the process  that the end result is not that convincing.    So, is that it? Is the hairstyle synthesis  dream dead? Well, if you have been holding   on to your papers, now squeeze that paper, and  have a look at this new technique. And, wow!    My goodness, just look at that. This is a huge  improvement just one more paper down the line.    And note that these previous methods  were not some ancient techniques,   these are from just one and two years ago. So  much improvement in so little time. I love it. Now let’s see what more it can do. Look… we  can choose the hairstyle we are looking for,   and, yes! It doesn’t just give us just one image  to work with, no! It gradually morphs our hair   into the desired target shape. That is fantastic,  because even if we don’t like the final hairstyle,   one of these intermediate images may give  us just what we are looking for. Loving it. And it has one more fantastic property. Look. Most  of the details of the face remain in the final   results. Phew! We don’t even have to morph into  a different person to be able to pull this off.    Now, interestingly, look,  the eye color may change,   but the rest seems to be very close to  the original. Not perfect, but very close. And, did you notice? Yes, it gets better!   Way better! We can even choose the structure,   this would be the hairstyle,  which can be one image,   and the appearance, this is more like the  hair color. So, does that mean that? Yes!    We can even choose them separately. That is,  take two hairstyles, and fuse them together. Now, have a look at this too. This one is an  exclusive example, to the best of my knowledge,   you can only see this here on Two Minute  Papers. A huge thank you for the authors   for creating these just for us. And I must  say that we could stop on any of these images,   and I am not sure if I would be able to tell  that it is synthetic. I may have a hunch due   to how flamboyant some of these hairstyles are,  but not due to quality concerns, that’s for sure. Even if we subject these  images to closer inspection,   the visual artifacts of the previous  techniques around the boundaries of the hair   seem to be completely gone. And I wonder  what this will be able to do just a couple   more papers down the line. What do you think? I’d  love to know, let me know in the comments below!

### Segment 2 (05:00 - 05:00) [5:00]

Thanks for watching and for your generous  support, and I'll see you next time!