Open Student Projects

## How To Apply Please read this section carefully. I might not answer you if you do not follow the instructions and I am very busy at the moment (or just reply with a link here). In order to apply write an email to zorah.laehner@uni-siegen.de with at least the following information: - a **short** introduction of yourself including your study program, semester and programming languages you have experience with - your transcript of records (this is the list of your courses and grades you can get from Unisono) - which project you are the most interested in (from the list below, or your own project but only 3D related topics) - some kind of acknowledgment that you are aware that I only give out projects related to 3D data and will not ask for a pure image based project (except it is in the list below) Optional but appreciated: - your CV - a description of experience you have with handling 3D data - names of people in the computer vision chair you have worked with before - any questions/constraints you have that you want to clear before doing the programming task I will then send you a small programming task related to 3D geometry processing (no prior knowledge required) which you have a week to solve. Afterwards we have a meeting where we will discuss your solution, I will ask some questions about Bachelor level math or data structure topics to make sure you have the needed background knowledge, and I give you a more detailed overview over the project(s). ## Available Projects Currently available spots: 0 (there are more topics here than available spots but I can only supervise a certain amount of students at the same time, please do not write emails if it says 0 spots available) All projects related to deep learning need to be implemented in PyTorch. ### Fast Marching on a 3D Product Manifold *Summary:* Replace the Dijkstra shortest path implementation in the paper below with fast marching. *Suitable for:* Bachelor Thesis, Studienarbeit *Most related paper:* [Efficient Globally Optimal 2D-to-3D Deformable Shape Matching, Lähner et al, 2016](https://zorah.github.io/publication/2016-cvpr-efficient-globally-optimal-2d-to-3d-deformable-shape-matching) *Requirements:* C++, Matlab (brief is fine, the Matlab part does not need to be changed) ### Analysis of PointNet Architecture *Summary:* Analysis of training strategies and extension of the PointNet architecture. *Suitable for:* Bachelor Thesis, Studienarbeit *Most related paper:* [PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation, Qi et al, 2017](https://arxiv.org/abs/1612.00593) *Requirements:* Python, preferably experience with either PyTorch or 3D geometry processing, for Studienarbeit: passed Deep Learning lecture offered in Siegen ### Generation of LEGO Models *Summary:* Automatic generation of buildable LEGO models from a given 3D mesh. This can be both learning based or procedural based on your preferences. *Suitable for:* Master Thesis *Most related paper:* [LEGO Builder: Automatic Generation of LEGO As- sembly Manual from 3D Polygon Model, Ono et al, 2013](https://asset-pdf.scinapse.io/prod/1974096800/1974096800.pdf) *Requirements:* Python, passed any lecture related to machine learning, computer vision, or computer graphics offered in Siegen ### Non-Rigid Puzzles *Summary:* Merging and correspondence calculation between multiple non-rigidly deformed shapes of the same object instance (e.g. humans in the same pose, see example paper). *Suitable for:* Master Thesis *Most related paper:* [Spectral Unions of Partial Deformable 3D Shapes, Moschella et al, 2021](https://arxiv.org/abs/2104.00514) *Requirements:* Python, passed any lecture related to machine learning offered in Siegen