Shape Analysis and 3D Deep Learning

---

Semester:	SS 2025
Type:	Lecture
Lecturer:	Dr. Isaak Lim
Credits:	ECTS 6 (V3/Ü2)
Contact:	shapeanalysis@cs.rwth-aachen.de

Course Dates:

Type	Date	Room
Lecture	Tue, 14:30-16:00	AH V
Lecture	Thu, 14:30-16:00	AH VI
Exercise	Thu, 12:30-14:00	AH III
Exam 1	August 19th, 9:00-10:30	H01
Exam 2	September 25th, 12:00-13:30	AachenMünchener Halle

In this lecture we will cover methodologies that allow us to analyse & extract higher-level information from single shapes or whole shape collections. This will allow us to tackle discriminative tasks such as segmentation or classification of shapes with little or large amounts of data. Beyond such tasks, we will also learn how to synthesise new data and generate 3D shapes.

To provide a robust foundation and enable you to make informed decisions, we will first focus on traditional (model-driven) methods. However, the core of this lecture will be on how to design neural networks that can take 3D shapes as input for various tasks. Towards the end of the lecture we will also cover more advanced topics and cutting-edge research.

Table of Content

1. Introduction
2. 3D Shape Representations
3. Distance Measures
4. Designed Descriptors
5. Designed Maps to Embedding Spaces
6. Model-driven Shape Analysis
7. Deep Learning Overview
8. Grid-based Approaches
9. Point-based Approaches
10. Graph-based Approaches
11. Shape Decoding
12. VQ-VAEs
13. GANs
14. Transformers
15. Diffusion
16. Differentiable Rendering & NeRFs
17. Aligning Modalities

Prerequisites

• Basic understanding of Neural Networks is recommended, but we will provide an "Introduction to Deep Learning"
• The lecture "Basic Techniques in Computer Graphics" and "Geometry Processing" is considered helpful, but not a hard requirement