The Second Workshop on
3D Human Understanding
A 2025 Perspective on 3D Human Perception, Reconstruction, and Synthesis
CVPR 2025
110B, Music City Center, Nashville
Embedded Files
Speakers
UC Berkeley
University of Tübingen
Seoul National University
MPI for Intelligent Systems
Meta
Overview
Overview
Understanding, reconstructing, and generating humans in 3D is a central challenge in computer vision, graphics, and robotics. This includes analyzing human pose and motion, modeling detailed appearance and clothing, synthesizing realistic behaviors, and capturing complex human-human interactions.
With the rise of large-scale data, generative models, and multimodal learning, the field now spans a wide range of topics—from pose estimation and clothed body reconstruction to neural rendering, motion synthesis, and LLM-based human understanding. These advances enable applications in AR/VR, digital avatars, human-robot collaboration, and beyond.
This workshop brings together researchers to explore the latest methods, highlight emerging challenges, and shape the future of 3D human understanding and generation.
The workshop will feature:
💡 Invited talks from leading voices in academia and industry
🖼️ Poster session showcasing work from recent top-tier conferences
🎙️ Panel discussion to explore open challenges and spark ideas for future research in 3D humans
Perceiving, Reconstructing, and Generating Humans
During social interactions
During social interactions
Under occlusions
Under occlusions
With fine-grained motion
With fine-grained motion
In the crowd
In the crowd
From egocentric views
From egocentric views
In 3D environments
In 3D environments
Schedule
13:50 - 14:00 Welcome and Opening Remark: Dimitris Tzionas
14:00 - 14:30 Talk #1: Angjoo Kanazawa
How to Train Your Humanoids From Video
14:30 - 15:00 Talk #2: Michael Black
Humans in Motion and Contact: Towards Understanding Human Behavior in 3D
15:00 - 15:30 Talk #3: Gerard Pons-Moll
How to Learn Large Scale 3D Human and Object/Scene Reconstruction Models
15:30 - 16:30 Poster Session (boards #315-334, ExHall D) + Coffee Break
16:30 - 17:00 Talk #4: Sebastian Starke
How to Break Your Neural Animation Controller
17:00 - 17:30 Talk #5: Hanbyul Joo
Learning 3D Human-Object Interactions from Generative Image Models
17:30 - 18:30 Panel discussion: Georgios Pavlakos + All Speakers
Organization Team
Organizers
NVIDIA
ETH Zürich
Meta
MPI for Intelligent Systems
Stanford
Meshcapade
Senior Advisory Board
Sponsor
If you are interested in serving as an additional sponsor, please contact stripathi@tue.mpg.de
Page updated
Google Sites
Report abuse