CVPR 2026

2nd Workshop on
Photorealistic 3D Head Avatars
(P3HA)


June 3rd 2026, 8:50am - 12:30pm

Colorado Convention Center, Denver CO, Room 107

Workshop Overview

Avatars are conquering the world! Let us explore the latest trends and techniques in research.
The Photorealistic 3D Head Avatars workshop takes place on June 3rd from 8:50 am - 12:30 pm in room 107.

Keynote Speakers

Workshop Program

08:50

Opening Remarks08:50 AM - 09:00 AM

Organizers

Organizers

Opening Remarks

09:00

Invited Talk 109:00 AM - 09:30 AM

Timo Bolkart

Semantic Correspondence: From Meshes to Gaussian Avatars
Bio

Timo Bolkart is a Senior Research Scientist at Google Zürich and former researcher at the Max Planck Institute for Intelligent Systems. His work lies at the intersection of computer vision, computer graphics, and machine learning, with a particular focus on 3D human face and body modeling, non-rigid shape analysis, and neural rendering. Timo has contributed to several influential projects in 3D vision and facial animation, including DECA, VOCA, and TEMPEH, and has authored numerous papers at top venues such as CVPR, SIGGRAPH, and ECCV.

09:30

Invited Talk 209:30 AM - 10:00 AM

Javier Romero

Are human representations already bitter enough?
Bio

Javier Romero is a Research Scientist at Meta's Codec Avatar research lab. He is a leading researcher in computer vision and 3D human modeling whose work has had a major impact on the fields of human body reconstruction, motion analysis, and digital humans. Javier is widely known as one of the creators of the influential SMPL body model, which has become a foundational representation for 3D human pose and shape estimation across computer vision and graphics research. His work bridges machine learning, geometry, and graphics, enabling realistic and scalable modeling of human motion and appearance for applications in AR/VR, animation, and embodied AI.

10:00

Invited Talk 310:00 AM - 10:30 AM

Juyong Zhang

Photo-realistic 3D Head Avatars: From Controllable Diffusion to Efficient Feed-Forward Inference
Bio

Juyong Zhang is a Professor at the University of Science and Technology of China (USTC) and a leading researcher in computer graphics and 3D vision. His research focuses on geometric modeling, digital humans, neural rendering, and high-fidelity 3D reconstruction, with influential contributions spanning computer vision, graphics, and AI-driven content generation. Juyong is a recipient of the Excellent Young Scholars Award from the National Science Foundation of China.

10:30

Benchmark Intro10:30 AM - 10:40 AM

Organizers

Organizers

Benchmark Introduction

10:40

Winner Talk 110:40 AM - 10:50 AM

URFace

URFace

Linzhou Li & Huakeng Ding

10:50

Winner Talk 210:50 AM - 11:00 AM

PMD-GAvatar

PMD-GAvatar

Kirill Chemrov

11:00

Invited Talk 411:00 AM - 11:30 AM

Vanessa Sklyarova

Modeling Strand-based Hairstyles for Digital Human Avatars
Bio

Vanessa Sklyarova is a PhD student at the Max Planck ETH Center for Learning Systems (MPI-IS & ETH Zürich) jointly supervised by Justus Thies (MPI), Michael Black (MPI), Otmar Hilliges (ETH) and Marc Pollefeys (ETH). Her work focuses on high-fidelity digital humans, with a particular emphasis on strand-level hair and fur reconstruction as well as neural rendering. She is a recipient of the Best Paper Runner-Up Award at 3DV 2026 for her work on NeuralFur, highlighting her contributions to realistic and physically grounded 3D reconstruction of complex natural structures.

11:30

Invited Talk 511:30 AM - 12:00 PM

Zhuo Su

Towards Deployable 3D Avatars: From Scalable Data Engine to Robust Modeling
Bio

Zhuo Su is a Tech Lead and Researcher at ByteDance, working on Human-Centric AI. His long-term goal is to develop machine intelligence grounded in human experiences, capable of understanding people, modeling human behaviors, and enabling natural interactions between humans, AI agents, robots and worlds. Zhuo has received several distinctions for both research and innovation, including the PICO “Star Team Award” Innovation Breakthrough Award at ByteDance, the Tencent Open Source Collaboration Award, and recognition as an Outstanding Graduate of Beijing and Tsinghua University.

12:00

Invited Talk 612:00 PM - 12:30 PM

Christian Theobalt

Highly Realistic Human Reconstruction and Rendering
Bio

Christian Theobalt is a director at the Max Planck Institute for Informatics and a leading figure in computer vision and computer graphics. His research has been instrumental in advancing neural rendering, performance capture, and digital human reconstruction, shaping many of the field’s most impactful developments. Christian is widely recognized for his scientific excellence, including being awarded an ERC Consolidator Grant (2017) and ERC Starting Grant (2013), the EUROGRAPHICS Outstanding Technical Contributions Award (2020), the EUROGRAPHICS Young Researcher Award (2009), and the German Pattern Recognition Award (DAGM, 2012). He is also a Fellow of EUROGRAPHICS and a recipient of the prestigious Otto Hahn Medal of the Max Planck Society, among numerous other honors recognizing both his research and leadership in the field.

Workshop Challenges

The workshop holds a competition on two tasks of the NeRSemble benchmark for 3D Head avatars. The goal is to find the current best method for single-view 3D face reconstruction and monocular FLAME-driven avatar creation.
Note that the dynamic novel view synthesis task is not part of this year's workshop challenges!

Single-view 3D Face Reconstruction

Given a single image of a person, the task is to produce a 3D mesh representing the person's head. There are 2 tracks:
  • Posed Reconstruction: The mesh needs to resemble the person and show the exact same facial expression.
  • Neutral Reconstruction: The mesh needs to resemble the person but have a completely neutral expression.
The challenge is conducted on 391 images from 20 different persons of various ethnicities, ages, and genders. The meshes for each image can be provided with FLAME topology or with arbitrary topology, in which case additional 7 landmarks are required for alignment.

Monocular FLAME Avatar Challenge (v2)

Given several frontal videos of a person's head with corresponding tracked meshes from FLAME, the task is to re-animate the person with unseen FLAME expression codes and then render from both seen (blue) and unseen (orange) camera viewpoints. This requires reconstructing an animatable 3D head representation (=3D head avatar). The challenge is conducted on recordings from 5 different individuals. For each individual, 18 short facial performance sequences are provided for training while the remaining 4 sequences are hold-out. For the hold-out sequences, only the tracked FLAME meshes and the camera poses are known.
The updated v2 of the benchmark task provides improved FLAME tracking to even better measure the reconstruction performance of avatar creation methods.

Competition Prizes

The winner of each workshop challenge will receive:
  • a dedicated 15-minute oral presentation in the workshop to showcase your method
  • an RTX 5080 GPU sponsored by NVIDIA*

*cannot be gifted to non-academics or persons residing outside of North America and Europe due to export restrictions imposed on NVIDIA by the US government

Competition Timeline

Date
Challenge begin 07th April 2026
Challenge submission deadline 26th May 2026
Winner announcement 28th May 2026

Workshop Organizers

Tobias Kirschstein Tobias Kirschstein Technical University of Munich
Simon Giebenhain Simon Giebenhain Technical University of Munich
Tianye Li Tianye Li NVIDIA
Koki Nagano Koki Nagano NVIDIA
Justus Thies Justus Thies Technical University of Darmstadt
Matthias Nießner Matthias Nießner Technical University of Munich

Workshop Sponsors

NVIDIA We thank NVIDIA for sponsoring the prices for the workshop challenge winners.

Please contact Tobias Kirschstein for questions.