1 of 25

3D Vision Foundation Models

Modern Methods for 3D Representation and Reconstruction

Joint Preliminary Meeting for Master Seminars

Shenhan Qian, Linus Härenstam-Nielsen, Ganlin Zhang, Weirong Chen

Computer Vision Group

08.07.2025

2 of 25

Outline

General Information

Organization
Registration

Example papers

3 of 25

Organization

The two seminars will be held jointly

Shared paper list
The only difference is the time slot

Slot 1: Wed 1-2 pm (3D Vision Foundation Models)
Slot 2: Wed 2-3 pm (Modern Methods for 3D Representation and Reconstruction)

Location: 02.09.023
Attendance to the chosen slot is mandatory
Attendance to the other slot is optional, but recommended

Weekly paper discussion

Each slot will have 1~2 presentations and discussion in English per week

4 of 25

Organization - Presenter

Give a 20 minutes presentation, followed by 10 minutes for Q&A session

Introduce the background, motivation, and related work
Explain the main findings and/or method of the paper
Summarize the experiments and how they support the paper claims
Give your own thoughts on the paper (pros and cons)

You may need to read related papers as well to fully understand the topic!
Send the slides to the instructors two weeks before the presentation for feedbacks

5 of 25

Organization - Audience

Before a seminar

Read the paper before the seminar
Submit a short summary (~150 words) of the papers one day before the seminar (Do NOT copy the abstract of the papers)

During a seminar

Listen carefully
Participate actively in Q&A

6 of 25

Organization - Reviewer

In addition to the selected paper to present, each student will also be assigned as reviewers of two papers, for which they need to prepare questions

Open the Q&A by asking one question each
Should read the paper before the presentation, and understand them well enough to ask meaningful questions

7 of 25

Assessment and Grading

Presenter (50%)

Paper Understanding

Clear explanation of key concepts, methods, and results
Highlighted motivations and contributions

Presentation Clarity and Structure

Well-organized and easy to follow
Slides clear, readable, and visually effective

Depth of Analysis

Beyond merely summarizing the paper
Well-discussed strengths and limitations

Audience (35%)

Papers summaries (15%)
Attendance (20%)
Active participation in Q&A (bonus)

Reviewer (15%)

Reviewing questions

8 of 25

Registration

Fill in the Google Form (link is also on our course website):

https://forms.gle/iPx4nSWTLqHMGPHg6

Applicants without filling the form will be ignored!

9 of 25

Example Papers

10 of 25

LRM: Large Reconstruction Model for Single Image to 3D

https://arxiv.org/abs/2311.04400

11 of 25

LRM: Large Reconstruction Model for Single Image to 3D

https://arxiv.org/abs/2311.04400

12 of 25

DUSt3R: Geometric 3D Vision Made Easy

https://arxiv.org/abs/2312.14132

13 of 25

DUSt3R: Geometric 3D Vision Made Easy

https://arxiv.org/abs/2312.14132

14 of 25

MASt3R-SLAM: Real-Time Dense SLAM with 3D Reconstruction Priors

https://arxiv.org/abs/2412.12392

15 of 25

MASt3R-SLAM: Real-Time Dense SLAM with 3D Reconstruction Priors

https://arxiv.org/abs/2412.12392

16 of 25

Dynamic Point Maps: A Versatile Representation for Dynamic 3D Reconstruction

https://arxiv.org/abs/2503.16318

17 of 25

Dynamic Point Maps: A Versatile Representation for Dynamic 3D Reconstruction

https://arxiv.org/abs/2503.16318

18 of 25

3D Gaussian Splatting for Real-Time Radiance Field Rendering

https://arxiv.org/pdf/2308.04079

19 of 25