iiinova — where your digital self keeps evolving

Reposted by Markus Worchel

Silvia Sellán @silviasellan.bsky.social · 19h

I'm one of the candidates for this year's Eurographics Executive Committee election! If you are an EG member, please make sure to vote :) My platform: - No AI in papers or reviews - Keeping what we like about Eurographics: the topics, the review quality, the scale. We won't become another ICLR

2152 15

Markus Worchel @mworchel.eurosky.social · 19/10/2026

May also risk desk rejection because it deviates too much from the given template. But not sure if anyone would *actually* enforce that in this situation..

000 0

Markus Worchel @mworchel.eurosky.social · 19/10/2026

(this is of course not on the authors but on the organizers) Also most conferences do have them in their submission template

000 0

Markus Worchel @mworchel.eurosky.social · 19/10/2026

Sure.. but why not simply add line numbers to the template? :D

100 1

Markus Worchel @mworchel.eurosky.social · 19/10/2026

How are reviewers supposed to refer to text or any part of a submission if the LaTeX template does not include line numbers? "On page 3, left side, two sentences after the definition of X ..."?

210 2

Reposted by Markus Worchel

Eurographics Conference @eurographicsc.bsky.social · 09/09/2026

📣 The Call for Full Papers for #Eurographics2027 is open! 📅 Abstract and submission form deadline: Sept. 25, 2026, at 23:59 UTC Discover the full call and submission details: 👉 eg2027.isti.cnr.it/call-for-pap... We look forward to receiving your contributions and welcoming you to #Lucca! #EG2027

043 4

Markus Worchel @mworchel.eurosky.social · 25/08/2026

Closely followed by my second favorite, "Burley and Studios (2012)"

020 2

Markus Worchel @mworchel.eurosky.social · 24/08/2026

Still delighted by how many papers refer to this as "Karis and Games (2013)"

1151 15

Reposted by Markus Worchel

Wenzel Jakob @wjakob.bsky.social · 17/08/2026

I'm excited to announce a first development release of nanobind 3.0. It introduces *split mode*, a new distribution model where a single Python wheel runs across all Python versions, while achieving outstanding performance and gaining rather than losing features! 🧵

1248 24

Reposted by Markus Worchel

Daniel J. Menezes @danjohnston.uk · 16/08/2026

Hey! Are you an indie dev who doesn't use gen AI and have a game you're currently promoting? I'm looking to make a starter pack and you could be in it! Reply below with a link to your game 👇 go.bsky.app/78u8M6Y

272676372 676

Markus Worchel @mworchel.eurosky.social · 31/07/2026

True, and that's not too far from reality. The most serious settings I know have some controlled environment for image-based object reconstruction...and the results are remarkable. But it's difficult to publish that because it mostly applies to one specific setup.

020 2

Markus Worchel @mworchel.eurosky.social · 30/07/2026

Even with a few hundred photos, there're often a gazillion possible solutions: Do we see a cloud of emitting particles (= NeRF), or surfaces reflecting light? We can restrict the solution space by choosing a smart parameterization.. and data-driven priors (un)fortunately help with that.

110 1

Markus Worchel @mworchel.eurosky.social · 30/07/2026

Nowadays, people are mainly exploring surfel-based representations (e.g. by retrofitting 3DGS to surface reconstruction), but my feeling is that the combination of implicits (topology/surface exploration) + mesh (local refinement) may be more promising.

010 1

Markus Worchel @mworchel.eurosky.social · 30/07/2026

A problem was runtime, which is one reason we explored using triangle meshes instead of implicit surfaces. It was faster, but we sacrificed accuracy, plus this required a sufficiently close initial guess to work properly (we used the visual hull). Meshes are great for local refinement, though.

210 2

Markus Worchel @mworchel.eurosky.social · 30/07/2026

Surely depends on your definition of "simplicity", but to me, the multi-view reconstruction techniques with differentiable rendering + implicit surfaces from ~2023 (IDR, NeuS, ...) always felt quite simple, and they worked pretty well (in object-centric settings).

100 1

Reposted by Markus Worchel

Jacco Bikker 🎦 @jbikker.bsky.social · 08/07/2026

TinyBVH has just been 𝘂𝗽𝗱𝗮𝘁𝗲𝗱 𝘁𝗼 𝟭.𝟴.𝟬. For CPU ray tracing, it is now the fastest option in most cases - Outperforming Intel Embree and Madmann91 for BVH build time and single-ray performance. On GPU, performance exceeds 4B rays per second - 4 rays per pixel at 4k/120Hz. github.com/jbikker/tiny...

68120 81

Markus Worchel @mworchel.eurosky.social · 03/07/2026

The code, including handle deformation and stylization demos, is available here: gitlab.com/tobidju/arap... Or, just run `pip install iarap` if you only want to use the regularization term. Thanks to @tobidju.bsky.social for leading the project + coauthors @ugoalone.bsky.social and Marc Alexa.

gitlab.com

Tobias Djuren / ARAP-Regularization-for-Implicit-Surfaces · GitLab

Source code for the paper 'As-Rigid-As-Possible Regularization for Implicit Surfaces' at SGP 2026.

031 3

Markus Worchel @mworchel.eurosky.social · 03/07/2026

We have seen heuristic, ARAP-like regularizers in many more applications with implicit surfaces (e.g. shape space learning), so we believe this is a useful tool beyond modeling. If you are interested, feel free to check out the paper (presented at SGP this week): diglib.eg.org/items/56b582...

120 2

Markus Worchel @mworchel.eurosky.social · 03/07/2026

Mesh-based modeling often builds on natural deformations. Equipped with an ARAP regularizer, we can apply the same ideas to implicit surfaces. For example, Gauss Stylization [Kohlbrenner and Finnendahl et al., 2021] is an artistic technique that aligns surface normals with preference directions.

110 1

Markus Worchel @mworchel.eurosky.social · 03/07/2026

The regularization term is theoretically grounded, simple to implement, efficient to evaluate...and it works surprisingly well!

110 1

Markus Worchel @mworchel.eurosky.social · 03/07/2026

Instead, we follow the (arguably) straightforward way: we start from the *continuous* ARAP energy by Chao et al., and derive a regularization term that is really nothing more than a Monte Carlo estimate. This only requires point samples on the undeformed surface. No meshing involved!

110 1

Markus Worchel @mworchel.eurosky.social · 03/07/2026

Prior work encourages as-rigid-as-possible deformations by retrofitting deformation energies designed for meshes to implicit surfaces through discretized level sets. This is not only costly but also error-prone: if the topology is not faithfully captured, topologically distant parts become coupled.

110 1

Markus Worchel @mworchel.eurosky.social · 03/07/2026

Implicit surfaces are everywhere in vision and graphics applications that optimize shapes. If we adjust the surfaces through a "learnable" deformation field, how can we ensure they stay close to the undeformed state (e.g. a template)? Our latest paper offers a simple and effective ARAP regularizer🧵

1173 17

Reposted by Markus Worchel

Wenzel Jakob @wjakob.bsky.social · 26/06/2026

I am excited to announce Dr.Jit 1.4 and Mitsuba 3.9 which now feature GPU-accelerated compute/ray tracing on Apple Silicon. Significant optimizations in both projects accelerate tracing/code generation ~2x. They generate better code and have faster Python bindings.🧵

1318 31

Reposted by Markus Worchel

Yining Karl Li @yiningkarlli.bsky.social · 18/06/2026

This is a wildly ambitious pivot, and a lot of the pitch is wrapped up in Silicon Valley hoopla, but I'd rather that AI labs shoot for things like this as opposed to mass slop generation. This is the kind of project that has pretty unambiguous potential for good if successful.

371 7

Markus Worchel @mworchel.eurosky.social · 15/06/2026

Maybe you're pitching it to the wrong folks? In my experience, non-ML graphics research is very much alive.

010 1

Reposted by Markus Worchel

Tzu-Mao Li @tzumaoli.bsky.social · 10/06/2026

cseweb.ucsd.edu/~tzli/cool_r... I started to maintain a list of homepages of graphics academics who I think did cool work, are younger than me, and are not my current/past students. It's a living list so if I read any cool papers I'll add the authors to the list. I encourage you to do the same!

cseweb.ucsd.edu

3155 15

Markus Worchel @mworchel.eurosky.social · 15/06/2026

Cool idea and feeling honored :)! I really wished that people would put more useful information on their homepage again instead of only posting to social media (or, even worse, closed platforms like Discord).

010 1

Markus Worchel @mworchel.eurosky.social · 21/05/2026

... it should be fine if the image is close to what 3DGS produces. And even if it isn't, it might still do something useful (hoping that momentum saves the day.. or the gradients).

100 1

Markus Worchel @mworchel.eurosky.social · 21/05/2026

Congratulations, cool work! :) I see why it doesn't make sense to differentiate through your algorithm (z-buffer test) but did you try to simply swap the forward pass of 3DGS with your implementation? Compute the loss with yours and use 3DGS backward. Sure, forward and backward mismatch, but ...

100 1

Reposted by Markus Worchel

Shlomi Steinberg @ssteinberg.xyz · 10/05/2026

youtu.be/Xolb8DcrbLc The recorded EUROGRAPHICS 2026 talk for our wave tracing paper. Very honored that this paper received the Günter Enderle Best Paper Honorable Mention award. Collaboration with Matt Pharr (NVIDIA).

youtu.be

Wave Tracing: Generalizing The Path Integral To Wave Optics – EUROGAPHICS 2026 Talk

YouTube video by Shlomi Steinberg

0223 22

Reposted by Markus Worchel

Tzu-Mao Li @tzumaoli.bsky.social · 04/05/2026

For people who are attending Eurographics in the coming week: my student Trevor will talk about his latest work on making ReSTIR efficient when reusing from a large number of neighbors. The key is an unbiased estimation of the MIS weight. See the paper below. research.nvidia.com/labs/rtr/pub...

1256 25

Markus Worchel @mworchel.eurosky.social · 03/05/2026

Tobias Djuren's "Basis Networks: Learning basis functions for free-form triangulations" learns basis functions from mesh examples to improve the limit surface of traditional subdivision schemes. Presentation: Tuesday · 14:00-15:30 · Kino 5 Paper: diglib.eg.org/handle/10.11...

000 0

Markus Worchel @mworchel.eurosky.social · 03/05/2026

Hendrik Meyer's "Hierarchical Optimization of the As-Rigid-As-Possible Energy" introduces a multi-level optimization of the ARAP energy for efficient surface deformation. Presentation: Wednesday · 14:00-15:30 · Kino 4 Paper: diglib.eg.org/handle/10.11...

100 1

Markus Worchel @mworchel.eurosky.social · 03/05/2026

Maximilian Kohlbrenner's "Contouring Signed Distance Fields by Approximating Gradients" improves SDF contouring by approximating gradient information using a tessellation. Presentation: Wednesday · 14:00-15:30 · Kino 4 Paper: diglib.eg.org/handle/10.11...

110 1

Markus Worchel @mworchel.eurosky.social · 03/05/2026

I'm on my way to Aachen and excited about my first time at Eurographics! I won't be presenting, so nothing new on differentiable rendering from our group this time :D But we have three cool papers on geometry processing:

270 7

Markus Worchel @mworchel.eurosky.social · 29/04/2026

I agree that if it's about funding, there's no point in being overly restrictive. My point was that blindly labeling anything as "AI" is misleading and, at worst, harmful. Especially to non-experts, the term "intelligence" can give a false sense of capability (I've seen this happen too often).

030 3

Markus Worchel @mworchel.eurosky.social · 29/04/2026

You can very likely relate *any* optimization problem to some abstract ML concept. Take surface deformation in geometric modeling: you have a large model (deformation per high-res mesh vertex) matching a set of point constraints to infer the positions of the unconstrained vertices. Is this "AI"?

100 1

Markus Worchel @mworchel.eurosky.social · 28/04/2026

Because investors love "AI". Apparently, solving optimization problems using gradient descent counts as "AI" nowadays.

020 2

Reposted by Markus Worchel

Christoph Peters @momentsingraphics.bsky.social · 31/03/2026

The High-Performance Graphics 2026 abstract deadline is coming up tomorrow (Wednesday, April 1, 23:59, anywhere on Earth). If you want to submit a paper, submitting an abstract beforehand is mandatory. www.highperformancegraphics.org/2026/call-fo...

highperformancegraphics.org

High-Performance Graphics 2026 CFP

High-Performance Graphics 2026 Call for Participation

031 3

Reposted by Markus Worchel

ceruleite.bsky.social @ceruleite.bsky.social · 19/03/2026

Four weeks to the EGSR papers deadline! (I'm so excited to be on the Program Committee this year - cannot wait to see all the amazing papers submitted!) So - submit away! Important dates: Abstracts deadline: 4/8 Papers deadline: 4/15 Submit your work, and join in Bordeaux! egsr2026.inria.fr

2219 21

Reposted by Markus Worchel

Christoph Peters @momentsingraphics.bsky.social · 19/03/2026

The papers deadline for High-Performance Graphics 2026 is 3 weeks from now and the abstract deadline is 2 weeks from now (April 1 AoE). If you want to submit, but have concerns about traveling to LA, please contact the paper chairs for options. www.highperformancegraphics.org/2026/call-fo...

highperformancegraphics.org

High-Performance Graphics 2026 CFP

High-Performance Graphics 2026 Call for Participation

0127 12

Reposted by Markus Worchel

Tzu-Mao Li @tzumaoli.bsky.social · 03/02/2026

I have procrastinated on writing so much that I wrote an entire document on writing tips: cseweb.ucsd.edu/~tzli/writin... Probably not much is new but I find I still need to repeat the same things to my students regularly. Will update this document over time hopefully.

cseweb.ucsd.edu

Writing tips

2185 18

Reposted by Markus Worchel

Tzu-Mao Li @tzumaoli.bsky.social · 09/01/2026

cseweb.ucsd.edu/~tzli/novelt... I gave an internal talk at UCSD last year regarding "novelty" in computer science research. In it I "debunked" some of the myth people seem to have about what is good research in computer science these days. People seemed to like it, so I thought I should share.

cseweb.ucsd.edu

27625 76

Reposted by Markus Worchel

Justin Solomon @justinmsolomon.bsky.social · 06/01/2026

Announcing SGI 2026! Undergrads and MS students: Apply for 6 weeks of paid summer geometry processing research. No experience needed: 1 week tutorials + 5 weeks of projects. Mentors are top researchers in this emerging branch of graphics/computing/math. sgi.mit.edu

1128 12

Reposted by Markus Worchel

Alex Battaglia @dachsjaeger.bsky.social · 06/01/2026

The EU/Europe and all NATO Partners need to put an end to what the Trump admin is doing and planning to do. The Greenland people are sovereign and decide their future. If the U.S. attempts to militarily annex them that is tantamount to war and subjugation. www.bbc.com/news/article...

bbc.com

US discussing options to acquire Greenland, including use of military - White House

In a statement, the White House told the BBC that Trump had made clear that acquiring Greenland was a "national security priority".

1121332 213

Markus Worchel @mworchel.eurosky.social · 18/12/2025

Fantastic presentation, worth a watch!

010 1

Reposted by Markus Worchel

Tzu-Mao Li @tzumaoli.bsky.social · 15/12/2025

Check out our recent SIGGRAPH Asia paper on differentiating programs that integrates over discontinuous functions (yes, again), which got the best paper award! yashbelhe.github.io/asd/index.html The first author Yash is looking for a job yashbelhe.github.io so talk to us if you're interested.

1162 16

Reposted by Markus Worchel

Arseny Kapoulkine @zeux.io · 08/12/2025

After nine years of development, meshoptimizer has reached its first major version, 1.0! This release focuses on improvements in clusterization and simplification as well as stabilization. Here's a release announcement with more details on past, present and future; please RT! meshoptimizer.org/v1

meshoptimizer.org

🐇 meshoptimizer v1.0

Mesh optimization library that makes meshes smaller and faster to render

924975 249

Markus Worchel @mworchel.eurosky.social · 01/12/2025

Wolfgang Straßer's PhD thesis being cited for its one-page future work section really shows how unpredictable 'impact' can be in research.

020 2