Skip to main content
Our research experts

Annika Hagemann

Ensuring high-accuracy camera calibration for automated driving and robotics

What I find fascinating is that vision seems quite effortless for us as humans – but implementing it in terms of algorithms generates an extremely high level of complexity.
Annika Hagemann

I am a PhD student at the Bosch Computer Vision Lab, where my research is focused on camera calibration. Having studied physics with a focus on computational neuroscience, I then moved to the more applied field of computer vision. The goal of my research is to accurately model how cameras map the 3D world onto a 2D image. Such a model is needed in automated driving and robotics, where cameras are used to perceive the surroundings and to navigate safely.

Curriculum vitae

  1. PhD student at Karlsruhe Institute of Technology & Bosch computer vision lab
  2. Max Planck Institute for Dynamics and Self-Organization, neural systems theory group
  3. Master of Science in Physics at Georg-August University Göttingen

Selected publications

  • Inferring bias and uncertainty in camera calibration

    A. Hagemann et al. (2021)

    Inferring bias and uncertainty in camera calibration
    • Hagemann, A., Knorr, M., Janssen, H., & Stiller, C.
    • International Journal of Computer Vision, 1-16.
  • Modeling dynamic target deformation in camera calibration.

    A. Hagemann et al. (2021)

    Modeling dynamic target deformation in camera calibration.
    • Hagemann, A., Knorr, M., & Stiller, C.
    • Accepetd for publication at IEEE/CVF, WACV 2022.
  • Bias Detection and Prediction of Mapping Errors in Camera Calibration

    A. Hagemann et al. (2021)

    Bias Detection and Prediction of Mapping Errors in Camera Calibration
    • Hagemann, A., Knorr, M., Janssen, H., & Stiller, C.
    • In Pattern Recognition: DAGM GCPR 2020. Springer International Publishing. (Best Paper Award 2020)
  • Assessing criticality in pre-seizure single-neuron activity of human epileptic cortex

    A. Hagemann et al. (2021)

    Assessing criticality in pre-seizure single-neuron activity of human epileptic cortex
    • Hagemann, A., Wilting, J., Samimizad, B., Mormann, F., & Priesemann, V.
    • PLoS computational biology, 17(3), e1008773.

Interview with Annika Hagemann

Annika Hagemann

PhD student for computer vision research

Please tell us what fascinates you most about research.

Research in computer vision means finding novel, better algorithms to enable machines to "see" their surroundings. What I find fascinating is that vision seems quite effortless for us as humans, but implementing it in terms of algorithms generates an extremely high level of complexity. Trying to cope with these challenges and coming up with new solutions is what makes research exciting.

What makes research done at Bosch so special?

I really appreciate the combination of basic research and application focus. Most of our research has a clear practical purpose and it is motivating to know that your results can be applied by others in future. Beyond that, there are many possibilities to discuss ideas – within the PhD community, with other researchers, and with colleagues from business units, who can give you a wide range of perspectives on your research topic.

What research topics are you currently working on at Bosch?

My current research is focused on high-accuracy camera calibration. This kind of calibration is needed whenever cameras are applied, for instance, to measure distances to other objects (e.g. in automated driving). Although the basic ideas behind camera calibration are well-known, ensuring the required accuracy remains a challenge. Therefore, my colleagues and I develop algorithms to detect calibration errors and to ensure high-accuracy calibrations.

What are the biggest scientific challenges in your field of research?

One of the main challenges in future will be the development of reliable algorithms for camera self-calibration, where camera characteristics are estimated online during usage. This is important as camera characteristics can change over time, which could introduce safety-critical errors. Unlike classical calibration approaches, self-calibration cannot rely on known 3D structures. To cope with this challenge, we are currently exploring combinations of deep learning and classical, geometric constraints.

How do the results of your research become part of solutions "Invented for life"?

An accurate perception of the surroundings is essential not only for automated driving, but also for intelligent robots, with applications spanning many different areas of our lives. By ensuring high-accuracy calibrations, we contribute to the accuracy and the safety of these systems.

Get in touch with me

Annika Hagemann
PhD student for computer vision research