Lab Course / Praktikum: Advanced Topics in 3D Computer Vision

by Prof. Nassir Navab
Team: Benjamin Busam, Mahdi Saleh, Patrick Ruhkamp, HyunJun Jung, Yida Wang
Type: Master Practical Course Module IN2106
SWS: 6
Programs: Informatics (Master), Biomedical Computing (Master)
ECTS: 10 Credits

Preliminary Meeting ("Vorbesprechung"): Thursday, 04.02.2021 at 11:00 AM. Zoom Meeting
Further Meetings: See the schedule below. According to the pandemic situation, we will partly or fully have our meetings / lectures via zoom.

This course in TUMOnline
The course will be conducted in English.

Announcements

• !NEW! Projects have been assigned
• Please submit your project preference until Mo. 10.05.2020 23.59h Project Voting
• Course matching finished.
• Registration for the course started, please use the application form and matching system.
• Application Deadline: 16.02.2021
• Preliminary Meeting: 04.02.2021

Registration

For the application, please use the following form (Deadline : 16.02.2021): Application Form
Please also register in the TUM matching system for the course registration. Keep in mind that your chances to be assigned to the course increase if you give it a higher rank in your choices. For further details about how the matching system works and its schedule please check this website.
We select appropriate candidates based on their background, interests, and motivation.

General Content

This course focuses on different recent topics in 3D computer vision and its application. The first set of lectures on specific vision problems (such as 6D Pose Estimation, Structure from Motion, Deep Learning) lays the foundation for practical challenges carried out by the student individually. While working on these problems the student becomes familiar with these topics and decides on a direction for the group phase. In the group phase, a group of students is supported by coaches from both industry and academia. The group identifies a particular 3D vision problem in one of these areas and develops a prototype to solve it. The prototype solution is finally presented with a live demonstration.

Example Applications

Example applications to follow.

Prerequisites

• Basic knowledge with C++ and Python programming
• Experience with basic computer vision and machine learning concepts
• Enthusiasm to tackle challenging problems
• Passion to create something new with a team

Passing Requirements

• Active participation in all mandatory workshops and meetings
• Submission of all individual challenge solutions in time
• Final group presentation with live prototype demonstration

Schedule

The schedule is preliminary and still subject to change. Please check the course website regularly!

Date	Time	Location	Topic	Conducted by	Materials
04.02.2021	11.00 - 11:30	Zoom	Preliminary Meeting via Zoom pwd:at3dcv	Patrick Ruhkamp Benjamin Busam HyunJun Jung	Preliminary Meeting Slides
13.04.2021	10.30 - 12.00	Zoom	Lecture 1: Basic Tips and Knowledge for Deep 3D Vision	HyunJun Jung	Course Organization Lecture Slides Challenge 01 Instructions Challenge 01 Lecture Video
19.04.2021	23:59 CET	Upload	Solutions to Challenge 1	Students
20.04.2021	10.30 - 12.00	Zoom	Lecture 2: Sensing Depth with 3D Computer Vision	Benjamin Busam	Lecture Slides Challenge 02 Instructions Challenge 02 Lecture Video
26.04.2021	23:59 CET	Upload	Solutions to Challenge 2	Students
27.04.2021	10.30 - 12.00	Zoom	Lecture 3: Visual Self-Localization - Classical Methods and Learning-based Approaches	Patrick Ruhkamp	Lecture Slides Challenge 03 Lecture Video
03.05.2021	23:59 CET	Upload	Solutions to Challenge 3	Students
04.05.2021	10.30 - 12.00	Zoom	Lecture 4: 3D Deep Learning	Mahdi Saleh	Lecture Slides Challenge 04 Lecture Video Projects Intro Project Introductions
10.05.2021	23:59 CET	Upload	Solutions to Challenge 4	Students
15.05.2021	10.00 - 14.30	Zoom	Project Brainstorming Workshop (attendance mandatory)	Benjamin Busam Mahdi Saleh Patrick Ruhkamp HyunJun Jung Yida Wang	Presentation Template
15.06.2021	10.30 - 12.00	Zoom	Project Update Presentations (attendance mandatory)	Benjamin Busam Mahdi Saleh Patrick Ruhkamp HyunJun Jung Yida Wang
15.07.2021	19.00 - ca. 21.30	Zoom	Final Workshop (attendance mandatory)	Benjamin Busam Mahdi Saleh Patrick Ruhkamp HyunJun Jung Yida Wang

Project Proposals

Project ID	Title	Supervisor	Students	Materials
#1	“PriorDepth“: Unsupervised Depth Estimation from Sparse Spatial-Temporal Priors	Patrick	Konstantin, Halil, Rahul	Proposal
#2	Hierarchical Retrieval in 3D Scans using GNN	Shun-Cheng, Mahdi	Andrew, Anna, Güven	Proposal
#3	Point Cloud Completion in Implicit Volumes	Yida	Berk, Yu-Ju, Ömer	Proposal
#4	ToF? Fusion	Hyun Jun	Dominik, Siyuan, Yuxuan	Proposal
#5	Polarimetric Pose Prediction	Ben	Iuliia, Magdalena, Yitong	Proposal

Coding

• Online C++ API reference
• Python 3.6 documentation
• OpenCV Documentation
• PyTorch documentation
• CMake
• MS Visual Studio (TUM StudiSoft)

Versioning

• Github, a freeware Github client for Mac/Windows
• Github tutorial for beginners
• With TUM account: LRZ Gitlab