Announcements
09/19/2023: Welcome to the course!09/19/2023: Homework 0 will be released on Piazza, due 10/03/2023 23:59 PM
General Information
Times & PlacesMoWeFr 6:00PM - 6:50PM, Warren Lecture Hall
Course Staff| Name | Office Hours | Location | ||
|---|---|---|---|---|
| Instructor | Prof. Hao Su | haosu@ucsd.edu | Wed 2:00pm-3:00pm | CSE 4114 |
| Teaching Assistant | Yunhao Fang | yuf026@ucsd.edu | Thu 12:00pm-1:00pm | CSE B250A |
Objectives
This is a graduate-level course to teach foundational and state-of-the-art concepts and algorithms of using deep learning methods for understanding and synthesizing 3D geometric data. The knowledge are widely used in applicatons of computer graphics, computer vision, and machine learning.Prerequisites
- Skilled in linear algebra
- Familiar with Multi-Variable Calculus
- Familiar with Probability and Numerical Methods
- Strong programming skills (Linux toolchain, Python, Numpy, PyTorch)
Grading
- Homework 0 5%
- Homework 1 30%
- Homework 2 30%
- Homework 3 35%
- Extra credit for participation 5% (ask/answer questions in class, attend office hours)
- There will not be a final exam.
Syllabus
- Geometry Basics
- Concepts of Classical Geometry
- Geometric Transformations
- 3D Representations
- 3D Recognition
- 3D Backbone Networks for Various Representations
- 6D Pose Estimation
- 3D Detection and Segmentation
- Open World 3D Understanding
- 3D Reconstruction
- Neural Fields
- Multi-View Stereo Network
- Diffentiable Isosurface
- 3D Diffusion Models
- Learning-based Mesh Processing
- Functional 3D Understanding
- Part Understanding
- Affordance Understanding
- Human and Hand Pose Understanding