Maximilian Weiherer
maximilian dot weiherer at fau dot de

I am a PhD student at FAU Erlangen-Nürnberg, advised by Prof. Bernhard Egger. I am part of the Cognitive Computer Vision Lab at the Chair of Visual Computing.

My research interests are broadly centered around computer vision, computer graphics, and machine learning. I am particularly interested in neural scene representations and anything related (including classical SSMs and 3DMMs) that could be used to model the world we live in. I also enjoy transferring my research into real-world applications, especially in the medical domain.

I obtained my B.Sc. and M.Sc. in Computer Science from OTH Regensburg in 2019 and 2021, respectively. Before joining FAU, I worked as a Student Researcher at the Regensburg Medical Image Computing (ReMIC) Lab, where I was advised by Prof. Christoph Palm. I also spent some time at Fraunhofer IIS as part of the Image Processing and Medical Engineering Department.

Google Scholar  /  LinkedIn  /  ResearchGate  /  X

For a full publication list, see my Google Scholar profile. Representative papers are highlighted.

We use and theoretically analyze the family of Matérn kernels for tunable implicit surface reconstruction, demonstrating that they outperform recently employed arc-cosine kernels.

We systematically compare and analyze four different strategies of how to incorporate multi-modality into neural scene representations, focusing on thermal imaging.

We introduce a new method to compare SSMs by computing approximate intersections and set-theoretic differences between the allowable shape domains spanned by the models.

We present the first publicly available 3D statistical shape model of the female breast.

Improved previously developed metamaterials to imitate human soft tissue and presented an optically tracked 3D printed phantom of the whole hand.

Extending the retrospective shading correction method based on signal envelope estimation to color images using principal color components.

We propose new 3D-printable support-filled metamaterials to imitate human soft tissue.

Using the difference of two elliptical paraboloids as approximation of the Laplacian of Gaussian, we enable real-time image stitching of microscopic images.

We present zero2hero, an innovative system that automatically extracts embarrassingly simple equations from a given Latex document and makes them look extremely complex.

We blame car manufacturers by showing that camouflaged cars (a.k.a. Erlkönige) can be better 3D reconstructed than its uncovered counterparts (without disturbing patterns).