Our Rising Stars 2026 application campaign is currently open so we thought we would share some updates about our past rising stars 🌟 follow us in the next couple weeks to read more!
Learn more about the program and apply >> www.wigraph.org/events/2026-...
My understanding is that in eg. the physics simulation context the approximation setting is pretty universal, because even the best exact convex decomp of a complicated shape is impractically complex for fast simulation.
(Though certainly the exact case is important and interesting mathematically!)
Thanks! I suspect your understanding is correct, hopefully the text of the paper makes the sense in which this is approximate clear.
Two equivalent viewpoints:
- We emit a union of convex bodies, approximating the shape
- We split the shape into pieces, each of which is approximately convex
Computing a convex decomposition of a shape is a classically-hard geometry problem, yet essential for fast physics simulators.
Yuezhi found a way to accelerate it by training a large model!
My PhD student Abhishek Madan is getting ready to graduate and he’s looking for a postdoc or industry position. Checkout his website www.dgp.toronto.edu/~amadan/
He’s been an amazing student , basically does all his research by himself. Fantastic math and implementation acumen.
📢want to create realistic dynamic 3D worlds (>100 splats)?
my NVIDIA internship project, VoMP, is the first feed-forward approach turning surface geometry into volumetric sim-ready assets with real-world materials.
🌐Project: research.nvidia.com/labs/sil/pro...
📜Paper: arxiv.org/abs/2510.22975
The Spatial Intelligence Lab at NVIDIA (research.nvidia.com/labs/sil/) is looking for 2026 research interns! We do all kinds of cool work across graphics/vision, geometry, physics, & ML. Now is the time to apply & reach out!
nvidia.eightfold.ai/careers/job/... (not limited to Canada-only)
Code is now out! Try it for yourself here: github.com/abhimadan/st...
Also: this paper was recognized with a best paper award at SGP! Huge thanks to the organizers & congrats to the other awardees.
I was super lucky to work with Yousuf on this one, he's truly the mastermind behind it all!
Actually, Yousuf did a quick experiment which is related (though a different formulation), using @markgillespie64.bsky.social et al's Discrete Torsion Connection markjgillespie.com/Research/Dis.... You get fun spiraling log maps! (image attached)
Yeah! That diffused frame is "the most regular frame field in the sense of transport along geodesics from the source", so you get out a log map that is as-regular-as-possible, in the same sense.
You could definitely use another frame field, and you'd get "log maps" warped along that field.
💻 Website: www.yousufsoliman.com/projects/the...
📗 Paper: www.yousufsoliman.com/projects/dow...
🔬 Code (C++ library): geometry-central.net/surface/algo...
🐍 Code (python bindings): github.com/nmwsharp/pot...
(point cloud code not available yet, let us know if you're interested!)
We give two variants of the algorithm, and show use cases for many problems like averaging values on surfaces, decaling, and stroke-aligned parameterization. It even works on point clouds!
Instead of the usual VxV scalar Laplacian, or a 2Vx2V vector Laplacian, we build a 3Vx3V homogenous "affine" Laplacian! This Laplacian allows new algorithms for simpler and more accurate computation of the logarithmic map, since it captures rotation and translation at once.
Previously in "The Vector Heat Method", we computed log maps with short-time heat flow, via a vector-valued Laplace matrix rotating between adjacent vertex tangent spaces.
The big new idea is to rotate **and translate** vectors, by working homogenous coordinates.
Logarithmic maps are incredibly useful for algorithms on surfaces--they're local 2D coordinates centered at a given source.
Yousuf Soliman and I found a better way to compute log maps w/ fast short-time heat flow in "The Affine Heat Method" presented @ SGP2025 today! 🧵
Holding SIGGRAPH Asia 2026 in Malaysia is a slap in the face to the rights of LGBTQ+ people. Especially now, when underrepresented people need as much support as we can possibly give them ! Angry like me ? Sign this open letter to let them know. 🏳️⚧️🏳️🌈
docs.google.com/document/d/1...
Sampling points on an implicit surface is surprisingly tricky, but we know how to cast rays against implicit surfaces! There's a classic relationship between line-intersections and surface-sampling, which turns out to be quite useful for geometry processing.
Thank you! There's definitely a low-frequency bias when stochastic preconditioning is enabled, but we only use it for the first ~half of training, then train as-usual. The hypothesis is that the bias in the 1st half helps escape bad minima, then we fit high-freqs in the 2nd half. Coarse to fine!
My child’s doll and tools I captured as 3D Gaussians, turned digital with collisions and dynamics. We are getting closer to bridging the gap between the world we can touch and digital 3D. Experience the bleeding edge at #NVIDIA Kaolin hands-on lab, #CVPR2025! Wed, 8-noon. tinyurl.com/nv-kaolin-cv...
Check out Abhishek's research!
I was honestly surprised by this result: classic Barnes-Hut already builds a good spatial hierarchy for approximating kernel summations, but you can do even better by adding some stochastic sampling, for significant speedups on the GPU @ matching average error.
Ah yes absolutely. That's a great example, we totally should have cited it!
When we looked around we found mannnnnny various "coarse-to-fine" like schemes appearing in the context of particular problems or architectures. As you say, what most excited us here is having simple+general option.
Thank you for the kind words :) The technique is very much in-the-vein of lots of related ideas in ML, graphics, and elsewhere, but hopefully directly studying it & sharing is useful to the community!
We did not try it w/ the Gaussians in this project (we really focused on the "query an Eulerian field" setting, which is not quite how Gaussian rendering works).
There are some very cool projects doing related things in that setting:
- ubc-vision.github.io/3dgs-mcmc/
- diglib.eg.org/items/b8ace7...
Tagging @selenaling.bsky.social and @merlin.ninja, who are both on here it turns out! 😁
website: research.nvidia.com/labs/toronto...
arxiv: arxiv.org/abs/2505.20473
code: github.com/iszihan/stoc...
Kudos go to Selena Ling who is the lead author of this work, during her internship with us at NVIDIA. Reach out to Selena or myself if you have any questions!
Closing thought: In geometry, half our algorithms are "just" Laplacians/smoothness/heat flow under the hood. In ML, half our techniques are "just" adding noise in the right place. Unsurprisingly, these two tools work great together in this project. I think there's a lot more to do in this vein!
Geometric initialization is a commonly-used technique to accelerate SDF field fitting, yet it often results in disastrous artifacts for non-object centric scenes. Stochastic preconditioning also helps to avoid floaters both with and without geometric initialization.
Neural field training can be sensitive to changes to hyperparameters. Stochastic preconditioning makes training more robust to hyperparameter choices, shown here in a histogram of PSNRs from fitting preconditioned and non-preconditioned fields across a range of hyperparameters.
We argue that this is a quick and easy form of coarse-to-fine optimization, applicable to nearly any objective or field representation. It matches or outperforms custom designed polices and staged coarse-to-fine schemes.
