Thinesh Thiyakesan Ponbagavathi
PhD Candidate in Computer Science
Stuttgart, Germany
Email Scholar GitHub LinkedIn CV
About
My research focusses on parameter-efficient fine-tuning techniques to adapt vision foundation models for fine-grained video understanding, with a focus on group activity understanding, cross-view generalization, and human-robot interaction.
Currently, I am a PhD candidate in Computer Science at the University of Stuttgart, advised by Prof. Dr. Alina Roitberg, and a scholar at IMPRS-IS. I joined Amazon as an Applied Scientist Intern in Santa Clara, California. Previously, I worked at Robert Bosch GmbH as a working student and at Johnson Controls as a Lead Controls Engineer.
News
- Sep 2026Our work on Not All Layers Are Equal in Image-to-Video Transfer has been accepted to NeurIPS 2026.
- Aug 2026Joined Amazon as an Applied Scientist Intern in Santa Clara, California 🇺🇸
- Jul 2026Attended International Computer Vision Summer School (ICVSS 2026) in Sicily 🇮🇹
- Jun 2026Attended CVPR 2026 in Denver 🇺🇸
- Mar 2026Attended VISAPP 2026 in Marbella 🇪🇸
- Feb 2026Our work on Frequency adapters for fine-grained video understanding accepted to CVPR 2026.
- Feb 2026Our work on efficient decoders for vision foundation models in group activity detection accepted to the SAUAFG Workshop at CVPR 2026.
- Jan 2026Our work on Nearly symmetric actions and order-aware image-to-video probing accepted to ICRA 2026.
- Jan 2026Our work on Pruning and distillation for real-time worker activity recognition accepted to VISAPP 2026.
- Nov 2025Attended ITSC 2025 on the Gold Coast 🇦🇺
- Nov 2025Temporal masking for cross-view foundation model probing in driver monitoring named Best Student Paper Finalist 🏆 (Top 5) at ITSC 2025.
- Oct 2025Oral presentation at the HLRS Workshop on High Performance Computing in Science & Engineering, Stuttgart 🇩🇪
- Jul 2025Our work on temporal masking for cross-view foundation model probing in driver monitoring accepted to ITSC 2025.
- Feb 2025Joined the International Max Planck Research School for Intelligent Systems (IMPRS-IS).
- Jan 2024Started PhD at University of Stuttgart.
Publications
Frame2Freq: Spectral Adapters for Fine-Grained Video Understanding
@InProceedings{Ponbagavathi_2026_CVPR,
author = {Ponbagavathi, Thinesh Thiyakesan and Seibold, Constantin and Roitberg, Alina},
title = {Frame2Freq: Spectral Adapters for Fine-Grained Video Understanding},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2026},
pages = {24073-24083}
}
Order Matters: On Parameter-Efficient Image-to-Video Probing for Recognizing Nearly Symmetric Actions
@INPROCEEDINGS{ponbagavathi2025order,
author={Ponbagavathi, Thinesh Thiyakesan and Roitberg, Alina},
booktitle={2026 IEEE International Conference on Robotics and Automation (ICRA)},
title={Order Matters: On Parameter-Efficient Image-to-Video Probing for Recognizing Nearly Symmetric Actions},
year={2026},
volume={},
number={},
pages={17675-17682},
keywords={Modeling;Probes;Printing;Videos;Accuracy;Robots;Human-robot interaction;Conferences;Tuning;Foundation models},
doi={10.1109/ICRA57385.2026.11695733}}
Structured Relational Reasoning for Group Activity Assessment
@InProceedings{Ponbagavathi_2026_CVPRW,
author = {Ponbagavathi, Thinesh Thiyakesan and Yang, Chengzheng and Roitberg, Alina},
title = {Structured Relational Reasoning for Group Activity Assessment},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
month = {June},
year = {2026},
pages = {8931-8940}
}
PDNet: Exploring Pruning and Distillation for Real-Time Worker Activity Recognition in Industrial Assembly
@conference{visapp26,
author={Malek Baba and Thinesh Thiyakesan Ponbagavathi and Alina Roitberg},
title={PDNet: Exploring Pruning and Distillation for Real-Time Worker Activity Recognition in Industrial Assembly},
booktitle={Proceedings of the 21st International Conference on Computer Vision Theory and Applications - Volume 3: VISAPP},
year={2026},
pages={624-630},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0014477200004084},
isbn={978-989-758-804-4},
}
T-MASK: Temporal Masking for Probing Foundation Models across Camera Views in Driver Monitoring
@INPROCEEDINGS{T_MASK_Ponbagavathi,
author={Ponbagavathi, Thinesh Thiyakesan and Peng, Kunyu and Roitberg, Alina},
booktitle={2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC)},
title={T-Mask: Temporal Masking for Probing Foundation Models Across Camera Views in Driver Monitoring},
year={2025},
volume={},
number={},
pages={2927-2934},
keywords={Training;Accuracy;Sensitivity;Foundation models;Activity recognition;Benchmark testing;Cameras;Robustness;Monitoring;Videos;Driver Activity Recognition;Cross-view Generalization;Vision Foundation Models},
doi={10.1109/ITSC60802.2025.11423550}}
Education
2024 – Present
Ph.D. Computer Science
University of Stuttgart · IMPRS-IS
Advised by Prof. Dr. Alina Roitberg. Research on parameter-efficient adaptation of vision foundation models for fine-grained video understanding.
2021 – 2023
M.Sc. Electrical Engineering (Smart Systems)
University of Stuttgart
Thesis: 3D-Aware Urban Scene Generation, advised by Prof. Dr. Bin Yang.