Lightnews — Scholar-powered news

Markus Worchel

@mworchel.bsky.social

200 followers 150 following 34 posts

Computer Graphics PhD student at TU Berlin. (differentiable) rendering, inverse graphics, GPGPU mworchel.github.io

Posts Media Videos Starter Packs

Markus Worchel @mworchel.bsky.social · 27d

Thanks!

Markus Worchel @mworchel.bsky.social · 27d

Beautiful project! Do you have any recommendations for literature on relativistic ray tracing (and practical implementations)?

Markus Worchel @mworchel.bsky.social · Sep 3

Depends on the community you're looking for. I'm happy with the number of 'graphics researchers and programmers' here on Bluesky. I don't feel like I'm really missing out on anything since leaving twitter.

Reposted by Markus Worchel

Wenzel Jakob @wjakob.bsky.social · Aug 7

Dr.Jit+Mitsuba just added support for fused neural networks, hash grids, and function freezing to eliminate tracing overheads. This significantly accelerates optimization &realtime workloads and enables custom Instant NGP and neural material/radiosity/path guiding projects. What will you do with it?

Markus Worchel @mworchel.bsky.social · Aug 18

I found this to be a great reference: github.com/rgl-epfl/cho...
The kernels are precompiled to PTX and shipped with the PyPI package. On the user machine, they are JIT compiled using the CUDA driver API (dynamically loaded). However, this doesn't link any library, so might not be what you need.

GitHub - rgl-epfl/cholespy: An easily integrable Cholesky solver on CPU and GPU

An easily integrable Cholesky solver on CPU and GPU - rgl-epfl/cholespy

Markus Worchel @mworchel.bsky.social · Aug 12

Paper: dl.acm.org/doi/pdf/10.1145/3730900
Code (based on Mitsuba 3): github.com/tjueterb/mis...

Markus Worchel @mworchel.bsky.social · Aug 12

This is joint work between the computer graphics and audio communication groups at TU Berlin, with @ugoalone.bsky.social, @tjueterb.bsky.social, Daniel Wujecki, Fabian Brinkmann, Stefan Weinzierl, and Marc Alexa.

Markus Worchel @mworchel.bsky.social · Aug 12

We extend Radiative Backpropagation (by @merlin.ninja)/Path Replay Backpropagation (by @deliovicini.bsky.social), to time-dependent energy transport, enabling backpropagation in linear-time with constant-memory. Despite simulating time-dependent functions, we only need one additional scalar per ray!

Markus Worchel @mworchel.bsky.social · Aug 12

Our key observation is that this duality between acoustics and graphics lets us apply concepts from physically-based differentiable rendering to acoustics, enabling gradient-based optimization of acoustic objectives.

Markus Worchel @mworchel.bsky.social · Aug 12

The simulations assume that energy travels along rays and even the rendering equation is the same as in graphics! Almost, at least. Time is an additional parameter, because sound cannot be assumed to travel instantaneously.

Markus Worchel @mworchel.bsky.social · Aug 12

Room acoustics design, for example of concert halls, is currently a time-consuming process: engineers often semi-manually tweak a room's geometry and materials in long simulation loops to improve characteristics like clarity.

Markus Worchel @mworchel.bsky.social · Aug 12

Differentiable rendering has transformed graphics and 3D vision, but what about other fields? Our SIGGRAPH 2025 introduces misuka, the first fully-differentiable path tracer for acoustics.

Reposted by Markus Worchel

Wenzel Jakob @wjakob.bsky.social · Aug 7

Methods like NeRF and Gaussian Splats model the world as radioactive fog, rendered using alpha blending. This produces great results.. but are volumes the only way to get there?🤔 Our new SIGGRAPH'25 paper directly reconstructs surfaces without heuristics or regularizers.

Reposted by Markus Worchel

Christoph Peters @momentsingraphics.bsky.social · Aug 3

Physically-based differentiable rendering enables inverse rendering, but handling visibility is hard. Our SIGGRAPH
2025 paper uses quadrics to importance sample silhouette edges--outperforming all existing unidirectional differentiable path tracers.
momentsingraphics.de/Siggraph2025...

Reposted by Markus Worchel

EurIPS Conference @euripsconf.bsky.social · Jul 16

EurIPS is coming! 📣 Mark your calendar for Dec. 2-7, 2025 in Copenhagen 📅

EurIPS is a community-organized conference where you can present accepted NeurIPS 2025 papers, endorsed by @neuripsconf.bsky.social and @nordicair.bsky.social and is co-developed by @ellis.eu

eurips.cc

Reposted by Markus Worchel

Yining Karl Li @yiningkarlli.bsky.social · Jul 7

At SIGGRAPH 2025, Intel, Disney, & Chaos are doing a course on the nitty gritty details of implementing path guiding methods in 3 production renderers: Cycles, Hyperion, and Corona. I helped write the course notes: 80 pages of great stuff! Coming soon.

s2025.conference-schedule.org/presentation...

Markus Worchel @mworchel.bsky.social · Jun 25

Feel free to reach out to me at the conference or via mail.

Special thanks to @wjakob.bsky.social for his valuable feedback and for giving us the historical context for the original papers.

Paper: diglib.eg.org/items/67d3e0...
Code: github.com/mworchel/mit... (hopefully soon in Mitsuba 3)

3/3

GitHub - mworchel/mitsuba-missing-term

Contribute to mworchel/mitsuba-missing-term development by creating an account on GitHub.

Markus Worchel @mworchel.bsky.social · Jun 25

The reason lies deeper in the theory of radiative backpropagation (RB), the technique underlying these integrators: it was derived under the assumption of static geometry. We extend this theory to non-static geometry and show how the simplicity of RB can be retained in this case. 2/3

Markus Worchel @mworchel.bsky.social · Jun 25

If you work on physically-based differentiable rendering, take a look at our EGSR 2025 paper "Radiative Backpropagation with Non-Static Geometry".

TL;DR: Mitsuba 3's prb_reparam and prb_projective integrators are biased when parameters affect geometry, but this unrelated to discontinuities! 1/3

Reposted by Markus Worchel

Keenan Crane @keenancrane.bsky.social · Jun 20

Folks in the #SIGGRAPH community:

You may or may not be aware of the controversy around the next #SIGGRAPHAsia location, summarized here www.cs.toronto.edu/~jacobson/we...

If you're concerned consider signing this letter docs.google.com/document/d/1...
via this form
docs.google.com/forms/d/e/1F...

Reposted by Markus Worchel

Christoph Peters @momentsingraphics.bsky.social · Jun 15

Spherical harmonics are widely used in graphics. Ever wondered how to compute their derivatives? My HPG 2025 poster shows that it is basically free.
It also comes with a code generator for C, C++, Python, GLSL and HLSL to compute SH with derivatives.
momentsingraphics.de/HPGPoster202...

Derivatives of Spherical Harmonics

momentsingraphics.de

Reposted by Markus Worchel

Tzu-Mao Li @tzumaoli.bsky.social · Jun 14

Rendering nerds! Check out our latest work "Vector-Valued Monte Carlo Integration Using Ratio Control Variates" that has just gotten the best paper award at SIGGRAPH 2025. This paper presents a method that reduces variance of a wide range of rendering and diff. rendering tasks with negligible cost.

Reposted by Markus Worchel

Stephen Hill @selfshadow.com · Jun 10

The SIGGRAPH Physically Based Shading course is back this year: blog.selfshadow.com/publications...

SIGGRAPH 2025 Course: Physically Based Shading in Theory and Practice

blog.selfshadow.com

Markus Worchel @mworchel.bsky.social · Jun 4

What I wanted to say is that APIs like OpenGL (its functions, structs, etc.) have been designed for some purpose, which, however, is not "educate about the foundations of CG", so designing a custom API, tailored to what *you* want to teach, might be a better (albeit time-consuming) option.

Markus Worchel @mworchel.bsky.social · Jun 4

The question is, what one is trying to achieve. Quickly set up graphics application or teach? If the latter, teach an API or computer graphics? OpenGL (immediate mode or not) is not necessarily the best tool to teach the principles of CG.