Tobias Zirr

466 posts

Tobias Zirr

Tobias Zirr

@alphanew

Research@NVIDIA. Into CG/rendering, ML, ergonomic programming, high-perf SW design. Formerly Intel, demos, space warping. :) Opinions are my own. (Archived)

Katılım Haziran 2011
360 Takip Edilen630 Takipçiler
Tobias Zirr retweetledi
Stavros Diolatzis
Stavros Diolatzis@StavrosDiol·
Tomorrow (Tuesday) at SIGGRAPH and 11:15 am we will be demonstrating N-Dimensional Gaussian Fitting interactively and will be giving some insight into the applications and method. Join us at Exhibit Hall F for our demo!
English
3
12
76
6.6K
Tobias Zirr retweetledi
Stavros Diolatzis
Stavros Diolatzis@StavrosDiol·
We are finally releasing the code for NDG so you can train 6D Radiance Fields, 10D+ Global Illumination or use it in your own applications. We are aiming to improve the codebase over time so reach out in Denver or online! github.com/intel/ndg-fitt…
Stavros Diolatzis@StavrosDiol

Time to share our #SIGGRAPH2024 paper "N-Dimensional Gaussians for Fitting of High Dimensional Functions"! We show that N-Dimensional Gaussian mixtures can be optimized to handle high dimensional inputs like MLPs with a fraction of the training times. sdiolatz.info/ndg-fitting/

English
0
18
104
9.2K
Tobias Zirr retweetledi
Wenzel Jakob {deprecation notice}
Wenzel Jakob {deprecation notice}@wenzeljakob·
I just released nanobind v2.0.0 featuring significantly improved typing infrastructure. It can now automatically build reference-quality stubs for your extensions (enabling completion in VS Code, static type checking via MyPy/PyLance, etc.) #version-2-0-0-may-23-2024" target="_blank" rel="nofollow noopener">nanobind.readthedocs.io/en/latest/chan…
Wenzel Jakob {deprecation notice} tweet media
English
1
17
157
11.2K
Tobias Zirr retweetledi
Stavros Diolatzis
Stavros Diolatzis@StavrosDiol·
This allows us to train 6D parameterization for real world scenes to reconstruct complex effects like on this CD
English
2
1
19
1.3K
Tobias Zirr
Tobias Zirr@alphanew·
It was a joy to work with @StavrosDiol on this one, dissecting a wide spectrum of hybrid to classic representations to capture the essence of balanced adaptivity in higher-dimensional representations. The end result is surprisingly explicit and compact!
Stavros Diolatzis@StavrosDiol

Time to share our #SIGGRAPH2024 paper "N-Dimensional Gaussians for Fitting of High Dimensional Functions"! We show that N-Dimensional Gaussian mixtures can be optimized to handle high dimensional inputs like MLPs with a fraction of the training times. sdiolatz.info/ndg-fitting/

English
0
0
3
315
Tobias Zirr retweetledi
Weier Philippe
Weier Philippe@PhilippeWeier·
Ever wanted to load a large asset in your GPU path tracer, only to find it doesn't fit in memory? Then check out our #SIGGRAPH2024 paper “N-BVH: Neural ray queries with bounding volume hierarchies” (with Alexander Rath, @exppad, @iliyang, Philipp Slusallek and @boubek)
English
3
44
164
18.4K
Tobias Zirr retweetledi
Eric Heitz
Eric Heitz@eric_heitz·
In our latest @I3DCONF paper, we transform any rasterizer into a differentiable one with minimal engineering and zero external dependencies. Optimize game assets and volumes, train 3D Gaussian Splats, etc. directly within your existing engine. Compute shaders is all you need!
English
16
163
809
103.4K
Tobias Zirr retweetledi
Attila Áfra
Attila Áfra@attila_afra·
Intel Open Image Denoise 2.2.0 is out! It improves image quality and CPU performance, adds support for Intel Meteor Lake GPUs, Apple GPUs (via Metal) and all ARM64 CPUs, removes dependency on CUDA Runtime (switches to Driver API), and more! github.com/OpenImageDenoi… #IamIntel
English
1
36
131
14.4K
Tobias Zirr retweetledi
Dieter
Dieter@dieter_debaets·
I am now using (a very limited subset of) @ocornut dear imgui directly from HLSL :) This will allow so many more (and better!) debugging and stats!
Dieter tweet media
English
11
36
331
29.9K
Tobias Zirr retweetledi
Anton Kaplanyan
Anton Kaplanyan@Kaplanyan·
Intel Graphics Research is excited to share our contribution across SIGGRAPH, EGSR, HPG, and EG this summer. Low-energy path tracing, generative AI, neural graphics and more! intel.com/content/www/us… Amazing work of our researchers and collaborators.
English
4
62
217
23.7K
Tobias Zirr retweetledi
Christoph Peters
Christoph Peters@MomentsInCG·
@HPG_Conf If you never understood how Nanite manages to avoid cracks, you should take a look at the appendix of the paper. It describes the method rigorously, including details that were not clear to me at first, and provides a formal proof that it works. momentsingraphics.de/HPG2023.html
Christoph Peters tweet media
English
0
12
99
7.9K
Tobias Zirr retweetledi
Christoph Peters
Christoph Peters@MomentsInCG·
Nanite rasterizes virtualized geometry by selecting ~20 million triangles with appropriate LOD each frame. Carsten Benthin and I wrote a @HPG_Conf paper on how to build a BVH for these 🔺s instead. Since they are already clustered, the build is 10x faster! momentsingraphics.de/HPG2023.html
Christoph Peters tweet media
English
12
150
628
71.9K
Tobias Zirr retweetledi
Anis Benyoub
Anis Benyoub@Auzaiffe·
Did you know you could importance sample the GGX BRDF using spherical caps? Our latest paper shows how and leads to an even simpler and faster algorithm than that of @eric_heitz. Check it out online: ggx-research.github.io/publication/20…
English
2
41
128
18.5K
Tobias Zirr retweetledi
Eric Heitz
Eric Heitz@eric_heitz·
When @_Laurent and I started learning about diffusion models, we were puzzled by the amount of jargon and concepts. So, we derived a model from scratch with our own graphics-people intuitions. Simple derivation, simple implementation, SOTA quality. ggx-research.github.io/publication/20…
Eric Heitz tweet media
English
28
340
1.8K
442K
Tobias Zirr retweetledi
Anton Kaplanyan
Anton Kaplanyan@Kaplanyan·
GGX team simplifies the concept of diffusion models to make it acceccible to more people around the world. As a bonus, it makes a connection to well-known graphics concepts. Check our this new work as a good way to start your journey in diffusion models intel.com/content/www/us…
English
1
37
182
14.5K
Tobias Zirr retweetledi
Stavros Diolatzis
Stavros Diolatzis@StavrosDiol·
I can finally share our CGF paper MesoGAN! We train a 3D GAN with a novel representation to allow generation of arbitrary extent 3D textures. We use these textures with shell mapping to create the appearance of mesoscale materials like fur or grass. More: repo-sam.inria.fr/fungraph/mesog…
English
1
13
39
5.5K

Keşfet

@StavrosDiol @exppad @iliyang @boubek @I3DCONF @ocornut @HPG_Conf @eric_heitz