PhD Position in Generative AI and Bayesian Estimation D Human Motion

Fully Funded PhD Position in Generative AI and Bayesian Estimation for 3D Human Motion

We are offering a fully funded Ph.D. position for a highly motivated researcher interested in generative artificial intelligence, probabilistic machine learning, computer vision, and Bayesian state estimation
.

The successful candidate will develop new generative and uncertainty-aware methods for estimating human motion in three dimensions. A particular focus will be placed on models that respect the geometry of human joint rotations and represent complete poses directly on products of rotation manifolds.

Affiliation: The position is affiliated with the research group of Jun.

-Prof. Dr.

-Ing. Florian Pfaff at the Institute of Industrial Automation and Software Engineering (IAS), University of Stuttgart.

Role: Research Associate / Doctoral Researcher leading to a Ph.D. degree.

Research area: Generative AI and recursive Bayesian estimation of 3D human motion on rotation manifolds.

The Ph.D. project is situated at the intersection of generative artificial intelligence, probabilistic machine learning, Bayesian state estimation, geometric deep learning, and 3D computer vision. The broader objective is to investigate new methods for uncertainty-aware modeling and estimation of three-dimensional human motion. Human poses consist of multiple joint rotations and therefore have an inherently non-Euclidean geometric structure. This creates fundamental research questions concerning how complex pose and motion distributions can be learned, how uncertainty can be represented and propagated over time, and how information from visual observations can be integrated into sequential estimation methods.

The project provides a broad scientific framework rather than a fixed sequence of predefined tasks. The successful candidate will be encouraged to develop their own research questions and methodological focus within this framework, taking into account their background, interests, and the findings that emerge during the Ph.D.

• Generative modeling of human poses and motions.
• Diffusion models, flow matching, normalizing flows, or related generative approaches.
• Probabilistic modeling on rotation manifolds and other non-Euclidean spaces.
• Learning and representing motion dynamics and observation uncertainty.
• Recursive Bayesian estimation and uncertainty propagation.
• Geometric deep learning for articulated systems.
• Robust estimation from ambiguous, noisy, incomplete, or occluded observations.
• Applications in human motion capture, computer vision, robotics, or human–machine interaction.

These directions are intended as an orientation and are not a fixed list of required work packages. Alternative approaches and related research questions that contribute to the overall scientific goals of the project are explicitly welcome.

Applicants should have:

• A very good Master’s degree in Electrical or Electronic Engineering, Information Technology, Computer Science, Autonomous Systems, Mechatronics, Cybernetics, Mathematics, Physics, or a closely related discipline.
• A strong foundation in machine learning, probability theory, statistics, or Bayesian estimation.
• Proficiency in Python.
• Practical experience with PyTorch or JAX.
• The ability and motivation to work independently on mathematically and technically challenging research problems.
• Strong communication skills and the ability to present research findings clearly in writing and at scientific meetings.

Experience in one or more of the following areas is particularly desirable:

• Diffusion models, flow matching, normalizing flows, or other generative models.
• Three-dimensional computer vision, human pose estimation, motion capture, or SMPL-based human-body models.
• Graph neural networks, transformers, or geometric deep learning.
• Scientific software development and GPU-based model training.

Applicants are not expected to have prior experience in all these areas. Candidates with a strong background in either generative machine learning, probabilistic estimation, computer vision, or mathematical modeling are encouraged to apply.

• A fully funded, full-time doctoral researcher position.
• Employment at 100% TV-L E13
  , according to the applicable German public-sector salary scale.
• The opportunity to pursue a Ph.D. degree at the University of Stuttgart.
• A research topic at the intersection of generative AI, probabilistic machine learning, computer vision, robotics, and nonlinear state estimation.
• Access to the group’s computational infrastructure and GPU resources.
• Opportunities to publish and present results at leading international conferences in machine learning, computer vision, robotics, graphics, and control.
• The opportunity to contribute to and release open-source research software.
• A collaborative research environment with considerable freedom to develop and pursue original research ideas.

Stuttgart, Germany.

Full-time and…