Tweet
The FAIR Center at Stanford University invites you to join our next webinar, featuring Tian Tan from Stanford University.
DETAILS
Title: GaitDynamics: A Generative AI Model for Gait Analysis
Speaker: Tian Tan, PhD, Wu Tsai Human Performance Alliance, Stanford University
Time: September 23, 2026 at 9:00 AM Pacific Time
Registration: Click here to register
ABSTRACT
The dynamics of human gait, including kinematics, joint moments, and external forces, offer key insights into human health and performance. Conventional gait analysis requires human experiments and physics-based simulations to quantify these dynamics. However, the high costs associated with human experiments and simulations has confined the scale of research studies. Data-driven models offer low-cost alternatives to performing experience by learning from large-scale datasets. Using the open-source AddBiomechanics dataset, Dr. Tan developed a generative model for efficient gait analysis across diverse downstream tasks. In the first part of the webinar, Dr. Tan will cover the two primary components of the study, including:
In the second part of the webinar, Dr. Tan will lead a hands-on tutorial using Hugging Face demos of the GaitDynamics model. He will demonstrate how to investigate gait-parameter changes across different running cadences and how to predict ground reaction forces from flexible combinations of input kinematics.
Tan, T., Van Wouwe, T., Werling, K. F., Liu, C. K., Delp, S. L., Hicks, J. L., & Chaudhari, A. S. (2026). GaitDynamics: A generative foundation model for analyzing human walking and running. Nature Biomedical Engineering, 1-13.
DETAILS
Title: GaitDynamics: A Generative AI Model for Gait Analysis
Speaker: Tian Tan, PhD, Wu Tsai Human Performance Alliance, Stanford University
Time: September 23, 2026 at 9:00 AM Pacific Time
Registration: Click here to register
ABSTRACT
The dynamics of human gait, including kinematics, joint moments, and external forces, offer key insights into human health and performance. Conventional gait analysis requires human experiments and physics-based simulations to quantify these dynamics. However, the high costs associated with human experiments and simulations has confined the scale of research studies. Data-driven models offer low-cost alternatives to performing experience by learning from large-scale datasets. Using the open-source AddBiomechanics dataset, Dr. Tan developed a generative model for efficient gait analysis across diverse downstream tasks. In the first part of the webinar, Dr. Tan will cover the two primary components of the study, including:
- Generative Architecture: GaitDynamics uses a diffusion-based architecture that iteratively generates outputs with flexible combinations of inputs.
- Downstream Applications: The model’s versatility across tasks with diverse inputs and outputs, such as using kinematics to estimate ground reaction forces and using synthetic trunk sway angles to predict their influence on knee adduction moments.
In the second part of the webinar, Dr. Tan will lead a hands-on tutorial using Hugging Face demos of the GaitDynamics model. He will demonstrate how to investigate gait-parameter changes across different running cadences and how to predict ground reaction forces from flexible combinations of input kinematics.
Tan, T., Van Wouwe, T., Werling, K. F., Liu, C. K., Delp, S. L., Hicks, J. L., & Chaudhari, A. S. (2026). GaitDynamics: A generative foundation model for analyzing human walking and running. Nature Biomedical Engineering, 1-13.