Lab Geometry Processing (MA-INF 2226, 9CP)
The introductory meeting (for all courses of the visual computing department) will be on October 9th, 14:00 over zoom. Please write an email to kneiphof at cs.uni-bonn.de for the link.
Description
This lab course will offer projects related to recent topics in (3D) geometry processing. For most this will mean implementing an existing baseline or algorithm and then improving one aspect of it. The topics this semester are:
Sketch Animation with Diffusion Models
Current sketch animation techniques based on text prompts work well on sketches that are automatically generated by algorithms but often do not work well for sketches drawn by humans (as they tend to not preserve the shape well, lines might be wobbly etc.). The task in this projects is to improve this by using collaborative score distillation and to allow more fine grained editing of parts.
Supervisor: Caro Schmitz (cschmitz at cs.uni-bonn.de)
Sketch Pose to 3D Human Model
3D modeling is a non-intuitive task for non-expert users. Drawing stick figures on the other hand can be done almost everyone. The goal of this project is to combine stick figure drawings with prompts for animations and 3D humans models to make it an intuitive process. This involves solving difficulties with the sparsity and ambiguity of sketches and learning to ignore mistakes in the drawing.
Supervisor: Caro Schmitz (cschmitz at cs.uni-bonn.de)
Learn to Make it Stand
A desirable property of 3D-printed objects is that they can keep standing without additional support structures. This property can for example be simulated and optimized with Make it Stand before printing. The goal of this project is to find out if properties and deformations like this can be learned by a neural network to speed up the process and reduce the necessary user interaction.
Supervisor: Zorah Lähner (laehner at uni-bonn.de)
Requirements
- There are no formal course requirements.
- All projects will be implemented in Python and prior experience is highly recommended. There will be no introduction to Python and the projects will be hard to complete while learning the language.
- Prior experience with 3D geometric data is beneficial but not necessary.
Participation
If you want to participate, write an email to the projects supervisor until October 14th, 2024. The projects can be done in group up to 3 or 4 people (more people will increase the expectations on the amount of work done) during the semester. The findings of the projects will be presented at the end of the semester. Participation in the final presentations of the course is mandatory.