StyleGAN-T: Unlocking the Power of GANs for Fast Large-Scale Text-to-Image Synthesis
significantly improves over previous GANs and outperforms distilled diffusion models in terms of sample quality and speed
abs: arxiv.org/abs/2301.09515
project page: sites.google.com/view/stylegan-
Follow
Click to Follow K_S_Schwarz
Katja Schwarz
@K_S_Schwarz
I am a PhD student in Computer Vision and Machine Learning. I am passionate about generative modeling in 2D and 3D.
Joined February 2020
Katja Schwarz’s Tweets
Andreas is a great advisor, don't miss the chance to come and work with us! :)
Quote Tweet
I am hiring PhD students and PostDocs! If you like to join a great team to conduct curiosity-driven research on implicit neural 3D representations join us now! Flyer: cvlibs.net/shared/flyer_p
2
30
🎅 Today, we have an early Christmas present for you: SDF Studio. Building on the fantastic code, we have integrated various implicit surface reconstruction techniques in one common framework! autonomousvision.github.io/sdfstudio/
More algorithms and results coming soon..
6
165
590
Considering a PhD and interested in differentiable rendering, self-supervised representation learning in vision, and 3D scene representations? Apply to MIT and consider my research group!
scenerepresentations.org
Application under gradapply.mit.edu/eecs/apply/log!
2
37
244
Show this thread
Together with we wrote a nice high-level intro to our latest publication VoxGRAF: Fast 3D-Aware Image Synthesis with Sparse Voxel Grids machinelearningforscience.de/en/escaping-pl
I'll present the work tomorrow at from 11am-1pm CST #522
If you're there come by for a chat!
Quote Tweet
Moving from 2D images to 3D graphics! Our new algorithm can be trained with 2D images alone to generate #3DGraphics and requires at the same time lower computational cost than usually. Check out our latest #BlogPost by @K_S_Schwarz. machinelearningforscience.de/en/escaping-pl
Show this thread
5
28
Code for VoxGRAF is out! github.com/autonomousvisi
It features:
- A 3D variant of StyleGAN using dense and/or sparse convolutions
- CUDA kernels for fast volume rendering for rgb, alpha and depth
- Pretrained models from the paper and tools for making cool videos like this one
0:04
5.1K views
1
17
139
Did you ever want to create a 3D avatar of yourself using only an RGB video from your smartphone? Now you can with ARAH: Animatable Volume Rendering of Articulated Human SDFs. Joint work with Shaofei to appear at : neuralbodies.github.io/arah/
0:13
2.9K views
2
21
127
Excited to present KING @ ECCV 22 w/ NiklasH : Using differentiable kinematics, our method generates novel safety-critical driving scenarios which improve the robustness of imitation learning based self-driving agents! lasnik.github.io/king/
GIF
15
44
Excited to announce VoxGRAF w/ : Motivated by INGP, we bring good old voxels back to the game of 3D aware image synthesis! Cool result: Rendering from any view is super fast (5 ms) after generating a feature grid (200 ms). katjaschwarz.github.io/voxgraf/
0:04
4.5K views
25
149
We are hiring an industrial PhD student co-supervised by myself and at and to work on identifiable object-centric representations. We would love to see applications from communities underrepresented in AI!
Apply here: robustml.is.mpg.de/pages/applicat
33
119
We are grateful to receive the Longuet-Higgins Prize at CVPR 2022. Thank you.
13
151
On your way to ?
Drop by our (w/ and ) tutorial on Diffusion Models (Sunday morning)! We just uploaded preliminary slides at 2-tutorial-diffusion-models.github.io
Here’s another artistic, fully diffusion model-generated interpretation of what to expect.
#CVPR2022
4
38
129
For an interactive view of a generated sparse voxel grid, check out our project page katjaschwarz.github.io/voxgraf/
6
Show this thread
Sparse voxel grids have proven super useful for speeding up novel view synthesis. Inspired by this, our latest work uses a sparse voxel grid representation for fast and 3D-consistent generative modeling.
Paper: arxiv.org/abs/2206.07695
Project page: katjaschwarz.github.io/voxgraf/
0:04
13K views
6
53
272
Show this thread
We finally released all benchmarks for the KITTI-360 dataset and got our TPAMI paper on KITTI-360 accepted! Huge shout out to Yiyi Liao for her hard work on this. Is KITTI-360 relevant to you? Find out in her new blog post:
14
66
🎉 The first #CyberValley Start-up #Incubation Program was exciting, lots of fun and a big success. A huge thank you to all participants, coaches, jury members, partners, and guests, who made this experience so special!
read image description
ALT
read image description
ALT
read image description
ALT
read image description
ALT
10
22
Two awards to our team in single day - best pitch and best business model in the startup #incubation program! Thanks for this amazing encouragement. And what a wonderful team:
2
3
31
After 6 weeks, today was the final pitch event of the Start-Up #Incubation program. The whole program was an amazing experience! + I'm so proud of our team to win both best pitch and best business model!
3
7
51
Excited to share that StyleGAN-XL has been accepted to 2022 🎉🎉🎉
For the camera-ready version, we pushed performance on ImageNet even further and also achieve SOTA on FFHQ at 1024^2 pixels!
Paper: arxiv.org/abs/2202.00273
Code: github.com/autonomousvisi
0:30
16.1K views
9
87
500
Show this thread
Open Phd / postdoc position in collaboration with Amazon AWS. You will work with , and myself. Topic Robust Machine Learning through 3D representations, but it's flexible. If you are interested apply until June 15th: gpmintern@listserv.uni-tuebingen.de
2
29
103
Show this thread
Excited to present RegNeRF! It enables realistic view synthesis from as few as three input images.
For more information, code, data, etc, see: m-niemeyer.github.io/regnerf/index.
Great collaboration with Mehdi S. M. Sajjadi and Noha Radwan!
0:25
16.6K views
4
67
430
The Autonomous Vision Group offers research fellowships for Ukrainian PhDs/PostDocs/researchers at risk in CV/NLP/ML as well as software engineering positions for BSc/MSc students. Please contact us for details and forward this to anyone this might apply to. Thanks.
3
54
178
. Demo for Projected GANs Converge Faster, Pokemon now on Spaces
github: github.com/autonomousvisi
Huggingface Spaces: huggingface.co/spaces/autonom
1
23
123
Show this thread
Come visit me at the NeurIPS poster session today starting at 5:30 CET!
eventhosts.gather.town/app/hwS3N8iLkK
I'm presenting our work "On the Frequency Bias of Generative Models" with Yiyi Liao and Andreas Geiger
Also, check out our brand new video:
youtube.com/watch?v=jbcB-h
0:20
2K views
8
55
First pokemon and now this *-*
Quote Tweet
𝑼𝒏𝒊𝒗𝒆𝒓𝒔𝒆 generated by Projected GAN:
Code: github.com/autonomousvisi
@AutoVisionGroup @kashyap7x @Drugwin
0:18
1.4K views
7
Wow, GRAF won the scientific award! Thanks to Cyber Valley for organizing this amazing opportunity to share our work with industry! And thanks again to my great collaborators Yiyi Liao, and Andreas Geiger from !!
Quote Tweet
These are the winners of the #AIGameDev! #mlxar/@ben_mlxar have won the main ward. @K_S_Schwarz's team with Generative Radiance Fields (GRAF) have won the scientific award. And the audience award goes to... @Kinetix_studio! Congratulations to all nominees! #CV5 #CyberValleyShow this thread
2
3
48
The finals are live on youtube at 10am CET. Looking forward to seeing some of you there!
Quote Tweet
The wait has come to an end: today is the day of the award ceremony of #AIGameDev
Make sure to watch our live stream from 10–12 CET on youtube.com/watch?v=sXNIqF to find out who wins the main & the science awards and use your vote to determine the winner of the audience award 
Make sure to watch our live stream from 10–12 CET on youtube.com/watch?v=sXNIqF to find out who wins the main & the science awards and use your vote to determine the winner of the audience award 
4
StyleGAN3 on Pokemon?! Can't wait!!!
Quote Tweet
Official Code Release for "Projected GANs Converge Faster"! You can now generate your own Pokemon:
github.com/autonomousvisi
(StyleGAN3 support coming soon)
@kashyap7x @AutoVisionGroup @ak92501 @arankomatsuzaki
0:08
13.3K views
7
Quote Tweet
Third nominee for #AIGameDev is @K_S_Schwarz with Generative Radiance Fields (GRAF) for 3D-aware #imagesynthesis
#AI #artificialintelligence #gamingtechnology #CyberValley @uni_tue
Award Ceremony live stream Dec 2, 10–12 CET: aigame.dev
3
39
The Autonomous Vision Group is looking for a PhD student at the intersection of NLP and CV to help revolutionize scientific discovery in the age of exponential research growth. Please help share with potential candidates!
cvlibs.net/shared/phd_nlp
2
30
113
Do you ever wonder what's really going on in your GAN? Our NeurIPS'21 paper investigates the cause of frequency artifacts in generated images. Nice addition: we develop simple testbeds to analyze generators and discriminators individually.
arxiv.org/abs/2111.02447
0:05
2.9K views
Quote Tweet
Today's GANs suffer from high frequencies artifacts. While most works attribute these to the generator, other works point to the discriminator. We rigorously analyze both the generator and discriminator in toy settings to shed light onto this problem. arxiv.org/abs/2111.02447
1
23
124
Can we use explicit representations to model implicit surfaces? In Shape as Points we propose a differentiable Poisson solver which does exactly that. Lightweight. Interpretable. 10x faster than neural implicits (eg, Occupancy Networks).
github.com/autonomousvisi
5
28
165
One of our most surprising findings this year: Perceptual features in the discriminator dramatically improve GAN convergence and FID scores across SotA models, but it only works when adding deterministic random projections! We still don't know exactly why.
sites.google.com/view/projected
3
54
224
Show this thread
Happening NOW, join and have a beer with us to celebrate the last day of ICCV main conference 🍻 twitter.com/ChrisJReiser/s
This Tweet is unavailable.
1
12
Excited to present our project UNISURF at #ICCV2021 in Session 5A & 5B with and .
Paper, Video, code, and more:
moechsle.github.io/unisurf/
GIF
1
7
60
Show this thread
Thanks for sharing! Check out our project page xuchen-ethz.github.io/snarf/ for more details and animation results.
Our code and models are publically available: github.com/xuchen-ethz/sn
GIF
This Tweet is unavailable.
4
26
The preprint about our new KITTI-360 dataset and its benchmarks and challenges is now online! We hope many people will find this dataset useful and that it will push self-driving and research at the intersection of vision, robotics and learning.
1
12
57
What can 1000 NeRF models do for you? Render 1000 times faster than a single NeRF! Sounds strange? Read 's blog on KiloNeRF:
autonomousvision.github.io/kilonerf/
0:11
1.7K views
12
94
We are hiring! Apply now to our International Max Planck Research School for Intelligent Systems. We are looking for the generation of doctoral students who are curious, creative, and passionate about intelligent systems: imprs.is.mpg.de/application
1
69
150