|about 1 month ago||10 days ago|
|GNU General Public License v3.0 or later||GNU General Public License v3.0 or later|
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
1 project | reddit.com/r/AnimeART | 14 Nov 2021
Made using https://thisanimedoesnotexist.ai/ from StyleGan2. Most of the images on the site are kind of cursed.
For people using Tesla GPUs in DIY setups, here's an open-source cooler design
1 project | reddit.com/r/homelab | 3 Nov 2021
Hi! I am using two of them to train/use stylegan2 neural networks. The config that can produce images at 1024/1024 requires ~10gb VRAM so these cards were the cheapest way to get into the door.
Open source/commercially available face generator?
2 projects | reddit.com/r/learnpython | 29 Oct 2021
Check the license here, but it looks like anything based on stylegan2 can't be used commercially. Google brought me to this page which seems to also be based on stylegan, but I guess they are paying for a commercial license for it? So you could use it or maybe even reach out to nvidia directly.
What's... Amodio? That's me!
1 project | reddit.com/r/Jeopardy | 25 Oct 2021
Not Matt (I'm sure he'll have more details), but yep! Specifically NVIDIA's StyleGAN2 architecture. The source code is freely available, so anyone with a good NVIDIA GPU can use it.
Ask HN: What useful unknown website do you wish more people knew about?
3 projects | news.ycombinator.com | 18 Oct 2021
StyleGAN2 (Dec 2019) - Karras et al. and Nvidia. More: https://github.com/NVlabs/stylegan2)
2. https://www.deepl.com/translator (translate text in 20 languages including idioms and phrases)
3. https://remove.bg/ (remove any background)
4. regex101.com (self explanatory)
5. Photopea.com (a free web-based Photoshop alternative
6. https://tineye.com/ , http://fotoforensics.com/ Do you want to know if an image is shopped, cropped or otherwise altered? Using these two tools you've got a good chance of finding out. Tineye is reverse image search on steroids and foto forensics provides free image analysis tools:
7. https://thenounproject.com/ (Icons and Photos for everything)
AI generated image for "Kerbal Space Program".
2 projects | reddit.com/r/KerbalSpaceProgram | 14 Oct 2021
If you’d like to know more, here’s a video explaining the method and the paper: https://arxiv.org/abs/1912.04958
Pre-trained StyleGAN2 model
2 projects | reddit.com/r/learnmachinelearning | 27 Jul 2021
The implementation and trained models are available on the StyleGAN2 GitHub repo.
AI-generated art day 3: Princess Luna!
1 project | reddit.com/r/mylittlepony | 11 Jul 2021
If you haven’t seen the previous posts, this art was made possible by This Pony Does Not Exist, which is powered by Nvidia’s StyleGAN2 AI. You can read more about the AI here.
[D] Paper regarding all ML problems reducing to diff eq..?
1 project | reddit.com/r/MachineLearning | 8 Jul 2021
Neural SDEs are the continuous-time limit of recurrent networks with noise as input. For example, StyleGAN2 is of this form. You can train Neural SDEs as VAEs or as GANs.
[R] Alias-Free GAN
1 project | reddit.com/r/MachineLearning | 29 Jun 2021
Greater than 99% consensus on human caused climate change in the literature
1 project | news.ycombinator.com | 20 Oct 2021
> Take your example about StyleGAN vs BigGan, I assume once it became clear that the latter was superior to the former that likely resulted in changes to existing architectures that then found additional improvements. This change in consensus is what enables that and is a good thing.
Well, no. :) But I think the "no" is because of the uniqueness of ML rather than a "no" to your point in general. You might be right about other fields; I don't have experience there.
In ML, there an enormous number of techniques. Style mixing was presented as a core feature of StyleGAN (https://arxiv.org/abs/1812.04948) and was enabled by default in the codebase (https://github.com/NVlabs/stylegan/blob/03563d18a0cf8d67d897...).
So there's a lot of "inertia" -- for example, when StyleGAN 2 came out, style mixing was still the default (https://github.com/NVlabs/stylegan2/blob/f2f751cdc7f996e3138...).
I haven't had time to dig into StyleGAN 3, but I suspect that style mixing might still be enabled by default.
It wasn't until we did a detailed, methodical analysis side-by-side with BigGAN, specifically to answer the question "Why is BigGAN so much better for diverse datasets?" that, on a whim, I turned off style mixing and was astonished to see BigGAN type quality pop out of a StyleGAN type arch.
Discoveries like that usually go unnoticed, frankly because it's a lot of effort to write a paper specifically to say "Hey, if you're training StyleGAN, definitely turn off style mixing. It only seems to work well on faces."
However, if such a paper were to be written, and accepted into a peer-reviewed journal, then your original point would probably be valid. So I don't even know if it's worth writing all of this -- I just thought it'd be interesting to point out the "Well, not really" in this case. The knowledge ends up floating around on Twitter and Discord rather than being transmitted via scientific papers...
But, this all does tie in to your final point:
> Consensus is easily changed with the introduction of new data, faith hangs on no matter how much evidence is put forward that it's horseshit.
It's remarkably easy for old, accepted ideas to hang around. You'd think it'd just be a matter of "Run the experiment; experiment proves thing; thing becomes accepted." But in practice it's felt quite different...
The thing is, everything you're saying is true in general. As t approaches infinity, there tends to be more and more consensus about older ideas, like the existence of black holes, or the validity of laws like F=ma. So we should probably pay attention when there is 99% consensus on a particular topic.
But, for example, one reason I wouldn't want to publish a paper claiming style mixing was bad, is because it would contradict the results of Karras, who is famous. I'd better be very certain about my claim! So there's sometimes a reluctance to contradict the consensus, too, which ends up equivalent to "faith" in your example -- we have faith that famous scientists are correct. (They usually are.)
As a cherry on top, I'll just leave a link to Feynman's messenger lectures: https://www.youtube.com/watch?v=-kFOXP026eE&ab_channel=TalkR... ... the history of science is fascinating. I'd dreamt for years of becoming a scientist, but the actual experience turned out to be surprisingly different than what I thought it'd be. I love it though -- all these weird corner cases are the spice of life.
Fakemon generated by AI
1 project | reddit.com/r/fakemon | 28 Sep 2021
[1812.04948] A Style-Based Generator Architecture for Generative Adversarial Networks
3 projects | reddit.com/r/LatestInML | 28 Aug 20212 projects | reddit.com/r/Regressions | 3 Jun 2021
PDF link Landing page2 projects | reddit.com/r/Regressions | 3 Jun 2021
Innovative Technology NVIDIA StyleGAN2
5 projects | dev.to | 15 Jun 2021
Code: https://github.com/NVlabs/stylegan5 projects | dev.to | 15 Jun 2021
https://arxiv.org/abs/1812.04948 (A Style-Based Generator Architecture for Generative Adversarial Networks)
AI made this stormy, overgrown, flodded version of the leaked image, I thought I'd share.
1 project | reddit.com/r/Battlefield6 | 3 May 2021
if anyone wonders I used Style2Gan NVlabs/stylegan: StyleGAN - Official TensorFlow Implementation (github.com)
Technoalchemy - a short spiritual book combining GPT-3 text and art made with StyleGAN + other ML techniques
3 projects | reddit.com/r/MediaSynthesis | 18 Apr 2021
Check out the PDF here I've been calling this "the first spiritual guidebook written by an AI". The text was written almost entirely with GPT-3, minus the couple of paragraphs I wrote as prompt. The art was made with various tools - for instance, the cover was made in part with Aphantasia (CLIP+FFT), the title page is an old public domain photo colorized with DeOldify, the faces made with StyleGAN, both standalone and via Artbreeder. Let me know what you think - I have a couple bigger projects building off of the techniques I used, and I would appreciate any feedback!
1 project | reddit.com/r/egg_irl | 2 Apr 2021
(Every person here is fake - the faces on the ends and the transition between them were created using StyleGAN)
What are some alternatives?
pix2pix - Image-to-image translation with conditional adversarial nets
stylegan2-pytorch - Simplest working implementation of Stylegan2, state of the art generative adversarial network, in Pytorch. Enabling everyone to experience disentanglement
awesome-pretrained-stylegan2 - A collection of pre-trained StyleGAN 2 models to download
ffhq-dataset - Flickr-Faces-HQ Dataset (FFHQ)
DeOldify - A Deep Learning based project for colorizing and restoring old images (and video!)
aphantasia - CLIP + FFT/DWT/RGB = text to image/video
stylegan2-ada - StyleGAN2 with adaptive discriminator augmentation (ADA) - Official TensorFlow implementation
waifu2x - Image Super-Resolution for Anime-Style Art