Picture of Soubarna Banik

Soubarna Banik, M.Sc.

Technical University of Munich

Postal address

Boltzmannstr. 3
85748 Garching b. München

Curriculum Vitae

  • (2019 - Present) Ph.D. candidate, Technische Universität München (TUM), Germany
  • (2018) Master of Science, Intelligent Adaptive Systems, University of Hamburg, Germany
  • (2009) Bachelor of Technology, Computer Science & Engineering, West Bengal University of Technology, India

Research Interests

  • Human posture detection and tracking
  • Multi object tracking
  • Graph convolutional networks
  • Deep learning
  • Visual attention


  • reFit: Research and development of an algorithm for posture recognition and tracking

Open thesis topics

If you are interested in any of the following topics, please send me an email mentioning your interest area and a list of related work that you have done before. 

  • human 2D to 3D pose tracking using Graph CNN and TRANSFORMER
  • human 2D and 3D pose estimation using Graph CNN
  • pose estimation in occluded scenes involving wheel chairs
  • head pose estimation for autonomous vehicles
  • Others: tracking, spatio-temporal data processing and 3D object detection

Bachelor thesis topics:

  • Multi-Object tracking for annotating markers of a motion capture system.

You will have opportunities to work with depth cameras, motion-capture system and deep learning.

Thesis Supervision


  • Bachelor thesis: Human Pose Estimation: Benchmark Dataset and Evaluation, Burak Ayaz, Oct, 2022 -
  • Master thesis: Tracking with Deep Learning Instance Segmentations, Ewva Ahmed, Oct, 2022 -
  • Master thesis: Evaluating the applicability of gamification to exercises based on the quadruped stand, Mark Pilgram, Oct, 2022 - 
  • Master thesis: Exploiting spatio-temporal relationships for occlusion-robust 3D human pose estimation, Patricia Gschossmann, Jun, 2022 -


  • Guided research: Joint rotation prediction for 3D human pose estimation, Apr, 2022 - Oct, 2022
  • Master thesis: Development and validation of a markerset to measure body movements in wheelchair children, Christin Bauer, Oct, 2021 - May, 2022
  • Bachelor thesis: Multi-object-tracking for annotating markers for a motion capture system, Kristiyan Sakalyan, Nov, 2021 - Mar, 2022
  • Master thesis: Evaluation of Gamification Aspects of a Vibration Board in Movement Therapy on the Example of Exercise Games, Hendrik Möller, Aug, 2021 - Feb, 2022
  • Master thesis: Graph Networks for Relational Inductive Bias in Vision-Based Deep Reinforcement Learning of Robotic Manipulator Control, Marco Oliva, Mar, 2021 - Sept, 2021
  • Master thesis: Development of a Long Term Motivation and Progression System in Serious Games for Movement Therapy, Daniel Hook, May. 2020 - Jan, 2021
  • Master thesis: Comparative Analysis of the State-of-the-art Image Synthesis Methods used for Object Classification in Automotive Industry, Tobias Schuster, Dec. 2019 -May, 2020


Previous Semesters:

  • Understanding the Human Body Using Deep Learning-Based Approaches (WS 2021/2022)



  • Soubarna Banik; Patricia Gschoßmann; Alejandro Mendoza Garcia; Alois Knoll: Occlusion Robust 3D Human Pose Estimation with StridedPoseGraphFormer and Data Augmentation, 2023 mehr… BibTeX


  • Oliva, Marco; Banik, Soubarna; Josifovski, Josip; Knoll, Alois: Graph Neural Networks for Relational Inductive Bias in Vision-based Deep Reinforcement Learning of Robot Control. IEEE Internatinal Joint Conference on Neural Networks, 2022 mehr… BibTeX Volltext (mediaTUM)


  • Banik, Soubarna; Garcia, Alejandro Mendoza; Kiwull, Lorenz; Berweck, Steffen; Knoll, Alois: Vogtareuth Rehab Depth Datasets: Benchmark for Marker-less Posture Estimation in Rehabilitation. 43rd Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), 2021 mehr… BibTeX Volltext (mediaTUM)
  • Banik, Soubarna; Gracia, Alejandro Mendoza; Knoll, Alois: 3D Human Pose Regression using Graph Convolutional Network. IEEE International Conference on Image Processing (ICIP), 2021 mehr… BibTeX Volltext (mediaTUM)
  • Banik, Soubarna; Lauri, Mikko; Knoll, Alois; Frintrop, Simone: Object Localization with Attribute Preference based on Top-Down Attention. 13th International Conference on Computer Vision Systems (ICVS), 2021 mehr… BibTeX Volltext (mediaTUM)