TU Wien Informatics

Overview Professors Scientific Staff Administrative Staff Student Staff External Lecturers By Responsibilities By Committee By Name
Related
Guest Professors Honoraries Alumni (LinkedIn) #5qw

Rafael Sterzinger

Univ.Ass. / BSc MSc ETH

Rafael Sterzinger

About

Research and Development Profile: Rafael Sterzinger

My research focuses on the intersection of Computer Vision and Digital Humanities, specifically exploring how Vision Foundation Models (VFMs) can be adapted for specialized, low-resource domains. The core objective is to bridge the gap between large-scale AI and the high-precision requirements of scientific analysis, where data is inherently scarce.

Core Research Interests
  • Cross-Domain Few-Shot Segmentation (CDFSS): Developing methodologies to generate precise pixel-level masks in visually disjoint target domains using only one to five examples.
  • Adaptive Model Architectures (MoE): Researching Mixture-of-Experts (MoE) strategies to dynamically merge specialized expert modules at inference time, allowing models to capture a broader spectrum of semantic concepts without extensive retraining.
  • Active Learning & Uncertainty Quantification: Leveraging MoE ensembles to reliably estimate predictive uncertainty, thereby optimizing human-in-the-loop workflows by identifying the most informative samples for expert annotation.
Scientific Contribution

My work addresses the fundamental challenges of domain shift, data scarcity, and high annotation costs in niche fields. By integrating model-centric adaptation with data-centric active learning, I aim to create optimized scientific workflows that minimize the need for expert labor while maximizing analytical capabilities for historical artifacts, such as ancient mirrors, manuscripts, and maps.

Role

Contact

2025W

2026S

 

  • Few-Shot Segmentation of Historical Maps via Linear Probing of Vision Foundation Models / Sterzinger, R., Peer, M., & Sablatnig, R. (2025). Few-Shot Segmentation of Historical Maps via Linear Probing of Vision Foundation Models. In X.-C. Yin, D. Karatzas, & D. Lopresti (Eds.), Document Analysis and Recognition – ICDAR 2025 : 19th International Conference  Wuhan, China, September 16–21, 2025  Proceedings, Part III (pp. 425–442). Springer. https://doi.org/10.1007/978-3-032-04624-6_25
  • Exploring Machine Learning for Faster Mapping and Scheduling of Automotive Applications on ADAS Platforms / Sterzinger, R., Koch, W., & Hoch, R. (2025). Exploring Machine Learning for Faster Mapping and Scheduling of Automotive Applications on ADAS Platforms. In M. A. Wani, P. Angelov, F. Luo, M. O. Wu Xintao, R.-E. Precup, R. Ramezani, & X. Gu (Eds.), 2024 International Conference on Machine Learning and Applications (ICMLA) (pp. 851–855). IEEE. https://doi.org/10.1109/ICMLA61862.2024.00123
  • Few-Shot Connectivity-Aware Text Line Segmentation in Historical Documents / Sterzinger, R., Lin, T., & Sablatnig, R. (2025). Few-Shot Connectivity-Aware Text Line Segmentation in Historical Documents. In Pattern Recognition and Computer Vision : 8th Asian Conference on Pattern Recognition, ACPR 2025, Gold Coast, QLD, Australia, November 10–13, 2025, Proceedings, Part I (pp. 116–130).
  • Segmentation of Historical Maps / Sterzinger, R. (2025). Segmentation of Historical Maps. In WORKSHOP ON SPATIAL DIGITAL HUMANITIES : BOOK OF ABSTRACTS AND CVs. WORKSHOP ON SPATIAL DIGITAL HUMANITIES, Wien, Austria.
  • Fusing Forces: Deep-Human-Guided Refinement of Segmentation Masks / Sterzinger, R., Stippel, C., & Sablatnig, R. (2025). Fusing Forces: Deep-Human-Guided Refinement of Segmentation Masks. In A. Antonacopoulos, S. Chaudhuri, R. Chellappa, Cl. Liu, S. Bhattacharya, & U. Pal (Eds.), Pattern Recognition (pp. 154–169). https://doi.org/10.1007/978-3-031-78198-8_11
  • Online HVAC Optimization under Comfort Constraints via Reinforcement Learning / Stippel, C., Sterzinger, R., Sengl, D., Bratukhin, A., Kobelrausch, M. D., Wilker, S., & Sauter, T. (2024). Online HVAC Optimization under Comfort Constraints via Reinforcement Learning. In IEEE Xplore (Ed.), 2024 IEEE 7th International Conference on Industrial Cyber-Physical Systems (ICPS). IEEE. https://doi.org/10.1109/ICPS59941.2024.10640003
  • Drawing the Line: Extracting Art from Ancient Etruscan Mirrors / Sterzinger, R. (2024, February 15). Drawing the Line: Extracting Art from Ancient Etruscan Mirrors [Conference Presentation]. COMPUTER VISION WINTER WORKSHOP 2024, Slovenia.
    Project: EtMirA (2021–2027)
  • Through-Wall Imaging Based On WiFi Channel State Information / Strohmayer, J., Sterzinger, R., Stippel, C., & Kampel, M. (2024). Through-Wall Imaging Based On WiFi Channel State Information. In 2024 IEEE International Conference on Image Processing (ICIP) (pp. 4000–4006). https://doi.org/10.1109/ICIP51287.2024.10647775
    Projects: Blindsight (2023–2025) / MAIJA (2024–2026)
  • Closing the Gap in Human Behavior Analysis: A Pipeline for Synthesizing Trimodal Data / Stippel, C., Heitzinger, T., Sterzinger, R., & Kampel, M. (2024). Closing the Gap in Human Behavior Analysis: A Pipeline for Synthesizing Trimodal Data. In 2024 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events (PerCom Workshops) (pp. 793–798). https://doi.org/10.1109/PerComWorkshops59983.2024.10503351
    Project: KIIS (2021–2023)
  • Drawing the Line: Deep Segmentation for Extracting Art from Ancient Etruscan Mirrors / Sterzinger, R., Brenner, S., & Sablatnig, R. (2024). Drawing the Line: Deep Segmentation for Extracting Art from Ancient Etruscan Mirrors. In Document Analysis and Recognition - ICDAR 2024 (pp. 39–56). https://doi.org/10.1007/978-3-031-70543-4_3
    Project: EtMirA (2021–2027)