PhysHuman @ CVPR 2026

Overview: Vision, graphics, and generative models can now reconstruct and synthesize humans with high visual fidelity. However, they rarely model how bodies should move under real-world physical constraints, such as contact, friction, joint limits, muscle effort, ground reaction forces (GRF), and center-of-mass (CoM) dynamics.

This workshop brings together computer vision, biomechanics, simulation, sports/rehabilitation, and XR researchers to make these physics quantities first-class targets for learning from video, IMU, and multimodal data. We will discuss datasets, metrics, and toolchains (e.g., OpenSim, MuJoCo, MyoSuite) that enable benchmarking of physical plausibility, and we will highlight applications in sports, clinical assessment, ergonomics, and safe human–digital interaction.

Topics covered in the workshop include but are not limited to:

• Vision-based estimation of physical quantities: GRF/CoM, joint torques/moments, and contact/friction states
• Physically grounded 3D human/body/face modeling, including musculoskeletal and soft-tissue models
• Modeling human together with wearables, exoskeletons, and footwear as a coupled physical system
• Physics-based human–object interaction, including contact reasoning, force estimation, and manipulation-aware motion understanding
• Physics-aware garment and material modeling for dynamic cloth–body interaction
• Physics-aware motion, pose, and avatar generation with joint limits and energetic priors
• Aligning simulations (OpenSim, MuJoCo, MyoSuite) with in-the-wild video, IMUs, and RGB-D data
• Datasets, metrics, and benchmarks for physical plausibility
• Applications in sports, clinical/rehabilitation assessment, ergonomics, and XR

We invite both short (up to 4 pages) and long (up to 8 pages) paper submissions, excluding references and supplementary materials. Submissions must follow the CVPR 2026 template. Authors should use the CVPR LaTeX style provided on the main website, available here. All papers will be subject to a double-blind review process.

• Full papers (archival): Up to 8 pages excluding references, for inclusion in CVPR 2026 workshop proceedings.
• Short papers (non-archival): Up to 4 pages for work-in-progress, negative results, or demos.

All accepted papers will be presented as posters, with selected papers featured as spotlight talks.