iiinova — where your digital self keeps evolving

Nicholas Sharp @nmwsharp.bsky.social · 18/08/2026

Excited to announce: In 2027 I'll join the UW Allen School in Seattle as tenure-track faculty in Computer Science! My group will advance core geometry processing + 3D AI for visual computing, science, & engineering. I'll be recruiting PhD students in the next cycle!

3411

Nicholas Sharp @nmwsharp.bsky.social · 05/08/2026

Tomorrow I'll speak at the FPTalks series on floating point and numerical analysis, about robust geometry for neural implicit surfaces (work w/ @_AlecJacobson). The other talks sound amazing, and I love this area! Check out this great organization. fptalks.org/talks/fptalk...

0132

Nicholas Sharp @nmwsharp.bsky.social · 30/07/2026

Have you dug deep into an evolving area of visual computing? Consider submitting to the 2027 EG STARs track! STARs are survey reports providing insight into the current state of the art and future directions. They're impactful and highly-cited! eg2027.isti.cnr.it/call-for-sta...

084

Nicholas Sharp @nmwsharp.bsky.social · 15/04/2026

Check out our Lyra2.0 for generating virtual worlds! The key is combining representations---the generative prior is a video model, but we leverage explicit 3D to scale to large scenes and support physical interaction. Huge kudos to Tianchang+Xuanchi who led it. research.nvidia.com/labs/sil/pro...

0101

Nicholas Sharp @nmwsharp.bsky.social · 02/07/2025

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)

010

Nicholas Sharp @nmwsharp.bsky.social · 02/07/2025

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!

180

Nicholas Sharp @nmwsharp.bsky.social · 02/07/2025

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.

160

Nicholas Sharp @nmwsharp.bsky.social · 02/07/2025

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.

160

Nicholas Sharp @nmwsharp.bsky.social · 02/07/2025

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! 🧵

26815

Nicholas Sharp @nmwsharp.bsky.social · 03/06/2025

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.

100

Nicholas Sharp @nmwsharp.bsky.social · 03/06/2025

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.

100

Nicholas Sharp @nmwsharp.bsky.social · 03/06/2025

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.

120

Nicholas Sharp @nmwsharp.bsky.social · 03/06/2025

Surprisingly, optimizing this blurred field to fit the objective greatly improves convergence, and in the end we anneal 𝛼 to 0 and are left with an ordinary un-blurred field.

100

Nicholas Sharp @nmwsharp.bsky.social · 03/06/2025

And implementing our method requires changing just a few lines of code!

100

Nicholas Sharp @nmwsharp.bsky.social · 03/06/2025

It’s as simple as perturbing query locations according to a normal distribution. This produces a stochastic estimate of the blurred neural field, with the level of blur proportional to a scale parameter 𝛼.

120

Nicholas Sharp @nmwsharp.bsky.social · 03/06/2025

Selena's #Siggraph25 work found a simple, nearly one-line change that greatly eases neural field optimization for a wide variety of existing representations. “Stochastic Preconditioning for Neural Field Optimization” by Selena Ling, Merlin Nimier-David, Alec Jacobson, & me.

3538