Jump to content

Talk:StyleGAN

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

February 2019

[edit]

This article is no more an advert than Deep Dream. kencf0618 (talk) 13:11, 16 February 2019 (UTC)[reply]

Then you'll be able to point us to substantial coverage of this website created yesterday. Wikipedia articles are used - as this one is - to buttress the credibility of someone's latest website. That's not what wikipedia is for. --Tagishsimon (talk) 14:27, 16 February 2019 (UTC)[reply]
The article is not about the web site per se, but GAN's state of the art. https://news.ycombinator.com/item?id=19144280 kencf0618 (talk) 19:42, 16 February 2019 (UTC)[reply]
Uh-huh. So starting the article with "This Person Does Not Exist (ThisPersonDoesNotExist.com) is a website..." is just there to mislead the children? Because correct me if I'm wrong, but we already have an article on Generative adversarial network. --Tagishsimon (talk)
No, and you are correct -otherwise I would not have linked to GAN. kencf0618 (talk) 03:11, 17 February 2019 (UTC)[reply]

Requested move 3 April 2019

[edit]
The following is a closed discussion of a requested move. Please do not modify it. Subsequent comments should be made in a new section on the talk page. Editors desiring to contest the closing decision should consider a move review after discussing it on the closer's talk page. No further edits should be made to this section.

The result of the move request was: Moved per consensus. No prejudice against nomination at proposed mergers per Safrolic. (closed by non-admin page mover) SITH (talk) 19:41, 27 April 2019 (UTC)[reply]



This Person Does Not ExistStyleGAN – This website is only one of many sites showcasing fully automated human image synthesis with StyleGAN and therefore it would be more beneficial for Wikipedia, its editors and readers to have info on the underlying algorithms and their various use cases instead of just promoting one of many sites where results obtained with StyleGAN are presented to the public. StyleGAN is not limited only to generating images that look like portraits of non-existent human faces unassisted, but could and likely will be used for much more sinister ends. Redress perhaps (talk) 12:28, 3 April 2019 (UTC) --Relisting. Dreamy Jazz 🎷 talk to me | my contributions 15:03, 10 April 2019 (UTC)--Relisting. B dash (talk) 03:40, 19 April 2019 (UTC)[reply]

  • Support. Nominator is correct that the underlying technology is likely much more notable than individual websites that use it — that's especially true if you have to rely on user-generated content sites like BoingBoing and Slashdot as your best references for the individual websites. Bearcat (talk) 20:40, 4 April 2019 (UTC)[reply]
  • Move suggester adds I have no training or knowledge of generative adversarial networks or AI other than some stuff I read on the Internets, so I kind of hard to write the article I would sincerely like to exist in English Wikipedia. And besides there's a more unnerving game version of presenting StyleGAN results than This Person Does Not, presents two pictures: a real human imaged with a real camera and another that it made by StyleGAN and you guess and the game tells you if you guessed right or not and if you'd like to play again: http://www.whichfaceisreal.com/ Redress perhaps (talk) 21:42, 8 April 2019 (UTC)[reply]
  • Comment The underlying technology is definitely more notable than the website itself, but it's not mentioned in any other context beside this site in this article, which might not really deserve to exist in the first place. Maybe redirect both of them to Human image synthesis and build out that article with the current state of different technologies first. Safrolic (talk) 02:43, 11 April 2019 (UTC)[reply]
Response to comment: Information about direct-to-2D synthesis based on neural networks, and specifically by generative adversarial networks, bypassing the old-skool 7D-route, definitely needs to be included/updated in relevant places. I'm busy doing things but I could look into updating human image synthesis at some point in time. Still think that StyleGAN is so shockingly good at what it does, that makes me worry what it would do if set to produce only images that look like someone. Redress perhaps (talk) 20:43, 13 April 2019 (UTC)[reply]

The above discussion is preserved as an archive of a requested move. Please do not modify it. Subsequent comments should be made in a new section on this talk page or in a move review. No further edits should be made to this section.

Architecture vs Implementation

[edit]

The sentence "StyleGAN depends on Nvidia's CUDA software, GPUs and on TensorFlow.[4]" is incorrect, it is also incorrect in the source. The StyleGAN architecture can be implemented on anything, it doesn't require Cuda, GPUs or TensorFlow. This only applies to the reference implementation provided with the paper. 2001:16B8:189C:5400:5547:4FE8:C5EA:70E9 (talk) 13:14, 24 October 2019 (UTC)[reply]

Not open source

[edit]

Stylegan cannot be used for commercial purposes, so it is not open source.--Mago Mercurio (talk) 13:49, 12 January 2020 (UTC)[reply]

Open source means the source code is available for inspection; it does not necessarily imply anything about licensing restrictions. 2A00:23C5:FE0C:2100:D443:E76E:29B1:9D77 (talk) 19:32, 26 July 2020 (UTC)[reply]