Introduction

Human beings are central to shaping and navigating the real world. As their digital counterparts, virtual humans are rapidly advancing into sophisticated replicas that reflect the complexity of real human existence. With significant strides in virtual human modeling benefiting areas such as autonomous systems, human-robot interaction, and immersive environments, a critical question arises: What truly makes a virtual human?

This workshop will investigate the key dimensions that define a virtual human, culminating the ultimate question of cognitive realism: (1) Human-Likeness: A virtual human must mirror not only the physical attributes but also the behavioral intricacies of real humans. We will explore the stateof-the-art in 3D avatar modeling, facial expressions, and motion capture technologies that contribute to achieving unparalleled visual and behavioral fidelity. (2) Interactivity and Agency: Virtual humans are interactive agents, capable of perceiving, making decisions, and acting autonomously in dynamic environments. We will address the challenges of enabling virtual humans to function autonomously, respond in real time, and adapt to changing conditions. Special attention will be given to advancements and bottlenecks in humanoid robots—focusing on better mimicking human behaviors, the role of visual awareness, and whether sophisticated modeling of human gait and adaptive control strategies are necessary for navigating complex terrains and performing human-like tasks. (3)Continuity and Adaptability: Like real humans, virtual humans must evolve over time. This discussion will focus on life-long modeling and learning frameworks for enabling virtual entities to grow, adapt, and maintain coherence. From a developmental aspect, these frameworks allow virtual humans to progressively change, acquire new skills and refine their abilities over time, ensuring they can continuously adapt to new tasks and environments as their capabilities evolve. (4)Ethics and Identity: As virtual humans integrate more deeply into real-world applications, the ethical implications surrounding their identity and usage must be addressed. We will discuss the principles of ethical design, the balance between autonomy and control, and how virtual humans will impact society.


The ability to replicate human-like appearance, behavior, and cognitive intelligence is the ultimate goal. Through advanced AI techniques, including virtual human modeling, reinforcement learning, and cognition modeling, we will discuss how virtual humans are evolving beyond visual realism to truly think, reason, and feel like real human beings—answering the fundamental question of what makes a virtual human.

News

Workshop website is launched.

Feb 03, 2025

POETS write: Hello CVPR'25!

Feb 03, 2025

Schedule

TBD.

TBD

Open Questions

Invited Speakers and Panelists


Steve Seitz

University of Washington

Bio
Siyu Tang

ETH Zürich

Bio
Jiajun Wu

Stanford University

Bio
Justus Thies

TU Darmstadt

Bio
Michael J. Black

Max Planck Institute for Intelligent Systems

Bio

Organiers


Kwan-Yee Lin

University of Michigan

Wayne Wu

University of California, Los Angeles

Bolei Zhou

University of California, Los Angeles

Matthias Nießner

Technical University of Munich

Stella X. Yu

University of Michigan