iiinova — where your digital self keeps evolving

Tamy Boubekeur

@tamyboubekeur.bsky.social

Compute graphics researcher, Senior Director of Adobe Research Paris, Professor at Ecole Polytechnique & Professor (on leave) at Telecom Paris. Views are mine. research.adobe.com/person/tamy-boub…

Posts Replies Media

Tamy Boubekeur @tamyboubekeur.bsky.social · 17/12/2025

Modeling faceted geometry as a 4D point process on a multiscale implicit grid enables real-time glint rendering for standard NDFs, with IS, anisotropy, and colored particles. Memory-free, in a standalone fragment shader. Our SIGGRAPH Asia 2025 paper: research.adobe.com/publication/....

0104

Reposted by Tamy Boubekeur

Élie Michel @elie-michel.bsky.social · 19/05/2025

🏅Honored to have been awarded at #Eurographics25 for our paper on #LipschitzPruning to speed-up SDF rendering! 👉 The paper's page: wbrbr.org/publications... Congrats to @wbrbr.bsky.social, M. Sanchez, @axelparis.bsky.social, T. Lambert, @tamyboubekeur.bsky.social, M. Paulin and T. Thonat!

0254

Tamy Boubekeur @tamyboubekeur.bsky.social · 26/06/2025

Triangles offer a natural strata to layout efficient non uniform random sampling on meshes. Our triangle rejection sampling reaches 10B samples/sec, which can feed primitive instancing, fiber distributions or emissivity sampling. Details in our HPG25 paper: research.adobe.com/publication/...

The surface mesh of a landscape (top left) is enriched with tens of millions of grass blades, designed in path traced in real time using a procedural density map (bottom left).

0257

Reposted by Tamy Boubekeur

Wilhem Barbier @wbrbr.bsky.social · 07/05/2025

I am proud to announce our Eurographics 2025 paper "Lipschitz Pruning: Hierarchical Simplification of Primitive-Based SDFs"! With Mathieu Sanchez (joint first author), @axelparis.bluesky.social, @elie-michel.bsky.social, Thibaud Lambert, @tamyboubekeur.bsky.social, Mathias Paulin and Théo Thonat.

Left: an input CSG tree and a much smaller pruned tree computed using our method.
Right: a rendered scene showing the number of active nodes per cell. Our method reduces the active nodes to less than 20 from the initial 6023 nodes of the input tree.

24618

Tamy Boubekeur @tamyboubekeur.bsky.social · 02/12/2024

ControlMat casts the "image-to-material" problem as a controlled generative task, guiding in-distribution channels diffusion with the input and introducing new diffusion components such as "Noise Rolling" for tileability. More in our ToG paper (presented Friday at SIG Asia): tinyurl.com/controlmat.

Two examples of photos (rock and leather) transformed into full PBR materials with ControlMat.

010