Gabor Fields: Orientation-Selective Level-of-Detail for Volume Rendering
Jorge Condor*, Nicolai Hermann*, Ata Yurtsever, Piotr Didyk
ACM Transactions on Graphics (to be presented at SIGGRAPH 2026)
Paper

Gabor Fields introduce a novel volume representation that supports continuous frequency filtering without extra memory overhead. By selectively pruning primitives and stochastically sampling frequencies and orientations, the method improves rendering performance, reduces aliasing, and enables efficient applications such as procedural cloud rendering.

Puzzle Similarity: A Perceptually-Guided Cross-Reference Metric for Artifact Detection in 3D Scene Reconstructions
Nicolai Hermann, Jorge Condor, Piotr Didyk
Proc. International Conference on Computer Vision (ICCV) 2025
Project Page Paper Code Data Supplemental Poster Demo

A novel no-reference metric to spatially assess 3D reconstructed view quality using perceptual embeddings, through an efficient patch-wise similarity computation between the training dataset and the evaluated view. Surpasses even direct reference metrics in human assessment correlation tests. We leverage it in automatic, recursive in-painting for artifact restoration.

Perceptually-Driven Neural Inpainting for Seamless 3D Reconstructions
Supervised by Prof. Piotr Didyk, Jorge Condor
Awarded the "Best Thesis Award" from the Premio Swiss Engineering Ticino foundation.
Thesis Announcement

In my master’s thesis, I improved novel Scene Reconstruction methods, such as Gaussian Splatting. I introduced a new approach that assesses reconstruction quality by leveraging the input multiview content as priors to evaluate novel views. I then demonstrated the method’s effectiveness by devising a strategy that hallucinates incorrectly reconstructed parts of a scene.

Reproducibility Challenge—PolyGCL: GRAPH CONTRASTIVE LEARNING via Learnable Spectral Polynomial Filters
Nicolai Hermann, Michele Cattaneo, Oliver Tryding
Project Report Reproduced Code WandB PolyGLC PolyGCL Code

In this paper, we replicate PolyGCL, which frames self-supervised graph contrastive learning as a spectral-polynomial fusion of high- and low-pass graph filters. Our study mostly reproduced the reported gains on both homophilic and heterophilic graphs—confirming its ability to learn expressive node embeddings—while noting that results still depend on careful hyper-parameter calibration relying on labeled data, compromising the self-supervised approach.

Robotics: Markov Localization and Monte Carlo Localization
Nicolai Hermann, Michele Cattaneo, Oliver Tryding
Project Report Code

In this project, we built a small simulator and implemented Markov localization to maintain a full probability grid of robot poses after each motion and sensor reading. For larger or continuous maps, we switched to Monte Carlo localization, replacing the exhaustive grid with a compact set of weighted particles (see blue points in the animation).

Avatar: The Legend of VR
Nicolai Hermann, Michael Hüppe, Hannah Köster, Jule Körner, Janeke Nemitz
Project Report Code

In this study, we explore whether users prefer interacting with virtual avatars that align with their visual preferences. In a VR experiment with 13 participants we found that avatar choices were influenced primarily by color and style. Results indicate a clear preference for visually congruent avatars.

Pokégans: Using Generative Adversarial Networks to Create the Next Generation of Pokémons
Nicolai Hermann, Michael Hüppe
Project Report Code

In this paper, we present a compact pipeline that trains a bespoke generative adversarial network to create convincing “Fakémon” sprites. Using a curated, uniformly pre-processed corpus of original Pokémon graphics, the model learns franchise-specific aesthetics and outputs novel, game-ready creatures. The paper details data acquisition and network design, showcases fan-project applications, and highlights future optimisation paths to boost fidelity and training efficiency.