Projects

ACM Transactions on Graphics · SIGGRAPH 2026

Forget Superresolution, Sample Adaptively (when Path Tracing)

Martin Bálint, Corentin Salaün, Hans-Peter Seidel, Karol Myszkowski

An end-to-end pipeline that learns where path-tracing samples matter, then reconstructs a detailed image from the sparse result.

Vision, Modeling, and Visualization 2025

Fizza Rubab, Ntumba Elie Nsampi, Martin Bálint, Felix Mujkanovic, Hans-Peter Seidel, Tobias Ritschel, Thomas Leimkühler

A learning-based treatment of antiderivatives using neural representations.

SIGGRAPH 2025 Conference Proceedings

Corentin Salaün, Martin Bálint, Laurent Belcour, Eric Heitz, Gurprit Singh, Karol Myszkowski

A stratification method for improving spatiotemporal reservoir sampling.

SIGGRAPH 2024 Conference Proceedings

Lingyan Ruan, Martin Bálint, Mojtaba Bemana, Krzysztof Wolski, Hans-Peter Seidel, Karol Myszkowski, Bin Chen

Self-supervised video defocus deblurring through an atlas-based representation.

SIGGRAPH Asia 2023 Conference Proceedings

Martin Bálint, Karol Myszkowski, Hans-Peter Seidel, Gurprit Singh

A method that treats sample allocation and stochastic optimization as one coupled inverse-rendering problem.

SIGGRAPH 2023 Conference Proceedings

Martin Bálint, Krzysztof Wolski, Karol Myszkowski, Hans-Peter Seidel, Rafał K. Mantiuk

A learned pyramidal filter that reconstructs low-sample-count path-traced images while avoiding ringing and scale-composition artifacts.

CESCG 2019 · Best Paper Award

Martin Bálint, Rafał K. Mantiuk

Closed-form transmittance for heterogeneous media represented with cosine noise.