What it is about
You will learn theoretical foundations and modern machine‑learning methods for 3D geometric data. The module covers representations of shapes and scenes as well as deep‑learning architectures for discriminative and generative tasks such as classification, segmentation, reconstruction and synthesis. In the end you will be able to understand and practically apply common approaches to point clouds, volumetric data, multi‑view inputs and graphs.
What you will be able to do
- Understand theoretical concepts of ML for 3D geometry
- Know deep‑learning architectures for discriminative and generative tasks
- Analyze models for different 3D representations (point clouds, graphs, volumes, multi‑view)
- Apply practical implementation methods and test on real tasks
What the module consists of
- LectureConveying the theoretical foundations of shape and scene analysis as well as deep‑learning architectures
- Exercises / HomeworkDeepening and practical application of the lecture material in the form of problems
- Final projectOwn small research project over the last two months for the practical demonstration of what has been learned (e.g., reconstruction from a single image, semantic segmentation)
Teaching method
- LectureExplanation of theory and discussion of practical applications
- Exercises (Homework)Deepening and practice of the concepts
- Project workPractical application and independent implementation of a research question as hands‑on experience