Rafael Sterzinger

Rafael Sterzinger is a PreDoc Researcher at the Institute of Visual Computing and Human-Centered Technology within TU Wien's Faculty of Informatics. Holding a BSc and MSc from ETH Zurich, he teaches courses in Computer Vision and Machine Learning for Visual Computing while pursuing his PhD.

Education

• MSc in Computer Science, ETH Zurich
• BSc in Computer Science, ETH Zurich

Research Interests

Sterzinger's work centers on Computer Vision and Visual Computing, specializing in segmentation techniques for historical documents, cultural heritage artifacts, and WiFi-based sensing applications. His research integrates human-in-the-loop systems with deep learning for refinement tasks, explores reinforcement learning for building automation, and develops domain adaptation methods for cross-scenario human activity recognition. Key domains include Etruscan art digitization, through-wall imaging, and historical map analysis.

Research Trends

Analysis of Sterzinger's 2023-2025 publications reveals three dominant threads: (1) Few-shot segmentation for historical materials using foundation models, (2) Human-guided refinement pipelines that merge algorithmic precision with user input, and (3) Wireless sensing applications leveraging WiFi channel state information for through-wall imaging and activity recognition. His work consistently addresses data scarcity through innovative adaptation techniques while maintaining real-world applicability in cultural heritage and smart infrastructure.

Projects

• EtMirA (2021-2027): Through-Wall Imaging Based on WiFi Channel State Information
• Blindsight (2023-2025)
• MAIJA (2024-2025): Closing the Gap in Human Behavior Analysis via Trimodal Data Synthesis
• KIIS (2021-2023): Deep Segmentation for Extracting Art from Ancient Etruscan Mirrors

Advising and Grants

No advisees are documented. Sterzinger participates in multiple externally funded projects including EtMirA's €1.2M infrastructure grant for cultural heritage digitization and MAIJA's interdisciplinary initiative synthesizing audio-visual-tactile data for behavior modeling, demonstrating sustained grant acquisition in visual computing domains.