This seminar covers advanced topic in digital humans with a focus on the latest research results. Topics include estimating human pose and motion from images, human motion synthesis, learning-based human avatar creation, learning neural implicit representations for humans, modeling, animations, artificial intelligence for digital characters, and others. A collection of research papers is selected.
Every participant has to present one of the papers presented in the first session. Additionally, you are required to read the paper that is presented in class beforehand and participate in a discussion during the seminar. An assistant will provide support when preparing the slides and in case technical questions arise.
The goal is to get an overview of actual research topics in the field of digital humans and to improve presentation and critical analysis skills.
Presence is mandatory to pass the seminar. If a student cannot attend a seminar session, the reason (e.g. medical certificate) has to be given before the session and must be accepted by one of the organizers. More than three missed seminar sessions will cause the student to fail this class. The dates for the presentations cannot be moved.
The presentation of the selected paper contributes 80% to the final grade. Additionally, the students are required to lead the discussion of another paper (20%).
| Number | 263-5702-00L |
| Lecturers | M. Gross, B. Solenthaler, O. Sorkine Hornung, S. Tang, R. Wampfler |
| Location | HG E 22, Thursdays 16:15-18:00 |
| Date | Paper | Presenter | Supervisor | Discussion |
| 10.10.24 | DNO: Optimizing Diffusion Noise Can Serve As Universal Motion Priors | Öykü Irmak Hatipoğlu | Korrawe Karunratanakul | Soroush Jahanzad |
| 10.10.24 | 3DGS-Avatar: Animatable Avatars via Deformable 3D Gaussian Splatting | Haonan Chen | Zhiyin Qian | Tianhao Chen |
| 17.10.24 | RoHM: Robust Human Motion Reconstruction via Diffusion | Tianhao Chen | Siwei Zhang | Yiming Wang |
| 24.10.24 | ResFields: Residual Neural Fields for Spatiotemporal Signals | Ruben Schenk | Marko Mihajlovic | Cedric Tschechtelin |
| 24.10.24 | SplatFields: Neural Gaussian Splats for Sparse 3D and 4D Reconstruction | Cedric Tschechtelin | Marko Mihajlovic | Ruben Schenk |
| 31.10.24 | DIMOS: Synthesizing Diverse Human Motions in 3D Indoor Scenes | Sihan Chen | Kaifeng Zhao | Öykü Irmak Hatipoğlu |
| 31.10.24 | Human3D: 3D Segmentation of Humans in Point Clouds with Synthetic Data | Soroush Jahanzad | Ayça Takmaz | Hairong Luo |
| 7.11.24 | Rendering with Style: Combining Traditional and Neural Approaches for High-Quality Face Rendering | Laura Nydegger | Yingyan Xu | Anna Banzer |
| 7.11.24 | Joker: Conditional 3D Head Synthesis with Extreme Facial Expressions | Mauro Vogel | Malte Prinzler | Laura Nydegger |
| 14.11.24 | Implicit Neural Representation for Physics-driven Actuated Soft Bodies | Shenghao Zhang | Lingchen Yang | Xichong Ling |
| 14.11.24 | SPARK: Self-supervised Personalized Real-time Monocular Face Capture | Hairong Luo | Christopher Otto | Haonan Chen |
| 21.11.24 | Learning a Generalized Physical Face Model From Data | Xichong Ling | Lingchen Yang | Shenghao Zhang |
| 21.11.24 | DreamFace: Progressive Generation of Animatable 3D Faces under Text Guidance | Rafael Fernandes | Yuxiao Zhou | Saatushan Sabesan |
| 28.11.24 | The Personality Dimensions GPT-3 Expresses During Human-Chatbot Interactions | Anna Banzer | Nikola Kovacevic | Maja Gwozdz |
| 28.11.24 | Rodin: A Generative Model for Sculpting 3D Digital Avatars Using Diffusion | Marc Matter | Manuel Kansy | Mauro Vogel |
| 5.12.24 | Example-based Motion Synthesis via Generative Motion Matching | Saatushan Sabesan | Peizhuo Li | Rafael Fernandes |
| 5.12.24 | Functional Maps: A Flexible Representation of Maps between Shapes | Yiming Wang | Jing Ren | Sihan Chen |
| 12.12.24 | Simplicits: Mesh-Free, Geometry-Agnostic, Elastic Simulation | Maja Gwozdz | Ruben Wiersma | Marc Matter |