Greetings! I’m Shao-Hsaun Hung, currently a PhD candidate at EEMCS,TU Delft. My research interests are primarily focused on (1) intersection of machine learning and signal processing and (2) deep learning for videos understanding (main on egocentric videos) . Previous, I am gaining research experience as an deep learning research intern at ASML Research, Data science and diagnosis team. My MSc thesis aims at a robust deep-learning based algorithm to automatically recognize (egocentric) procedural actions within an industrial setting. I just finished my MSc defense and graduated with GPA 3.97/4.0 in TU/e.
I also a member of Formula Student (student engineering competition that designing formula style racing car) through my undergraduate study. In my leisure time, I am a jogging hobbyist (I was a track & field athlete in high school!) and squash beginner.
My research interest covers the following aspects:
- Egocentric video understanding
- Robust vision models for real-world application
🔥 News
- 2025.10: 🎉 A paper got accepted by Elsevier Computer Vision and Image Understanding.
- 2025.09: 🎉 A paper got accepted by NeurIPS 2025 workshop.
- 2025.05: I will be a PhD candidate at EEMCS, TU Deflt, supervised by Prof. Raj Thilak Rajan, and Prof. Alle-Jan van der Veen.
- 2024.09: 🎉 A paper got accepted by IEEE Robotics and Automation letters.
- 2024.09: 🎉 I am graduated from TU Eindhoven with GPA 8.6/10, (3.97/4.0).
- 2024.08: 🎉 A paper got accepted by ECCV 2024.
- 2023.11: I will be an intern at ASML Research, supervised by Dr. Jacek Krustra (ASML Research), and Prof. Fons van der Sommen (TU/e).
- 2023.07: I will be an intern at TU Delft, supervised by Prof. Zhiyong Sun (TU/e), and Prof. Justine Dauwels (TU Delft).
📝 Publications
Shao-Hsuan Hung*, Tim J. Schoonbeek*, Dan Lehman, Hans Onvlee, Jacek Kustra, Peter H.N. de With, Fons van der Sommen,
- TLDR: STORM-PSR pairs a transformer-based spatio-temporal stream with assembly state cues and weakly supervised pretraining to recognize correctly completed steps even under partial occlusion.
BlockGPT: Spatio-Temporal Modelling of Rainfall via Frame-Level Autoregression
Cristian Meo*, Varun Sarathchandran*, Avijit Majhi*, Shao-Hsuan Hung, Carlo Saccardi, Ruben Imhoff, Roberto Deidda, Remko Uijlenhoet, Justin Dauwels Project page
- TLDR: BlockGPT is a spatio-temporal generative model designed for precipitation nowcasting. It models weather evolution as a video prediction task using transformer-based architectures.
Supervised Representation Learning towards Generalizable Assembly State Recognition
Tim J. Schoonbeek, Goutham Balachandran, Hans Onvlee, Tim Houben, Shao-Hsuan Hung, Jacek Kustra, Peter H.N. de With, Fons van der Sommen,
- TLDR: Using representation learning (rather than classification) for assembly state recognition beats classification approaches and improves generalization
Find the Assembly Mistakes: Error Segmentation for Industrial Applications
Dan Lehman*, Tim J. Schoonbeek*, Shao-Hsuan Hung, Jacek Kustra, Peter H.N. de With, Fons van der Sommen
- TLDR: Applying Change Detection Algorithms for Error Segmentation
🎓 Educations
- 2025.05 - Present, Ph.D. in Electrical Engineering, Delft University of Technology, the Netherlands Research field: Distributed signal processing, Gaussian Processes
- 2022.09 - 2024.09, MS in Electrical Engineering, Eindhoven University of Technology, the Netherlands
- 2016.09 - 2021.06, BS in Electrical Engineering, (dual degree), National Cheng Kung University, Taiwan
- 2016.09 - 2021.06, BS in Mechanical Engineering, (dual degree), National Cheng Kung University, Taiwan
👨💻 Internships
- 2025.06 - Now, Sensor AI Lab, EEMCS, TU Delft, Netherlands
- 2023.05 - 2020.02, ASML Research, the Netherlands
- 2023.05 - 2020.02, Justine Dauwels Lab, TU Delft, the Netherlands
- 2023.05 - 2020.02, Texas Instruments, Taiwan
- 2020.02 - 2021.08, System Dynamics Laboratory for Mechatronics and Microsystems Lab, Taiwan
- 2019.06 - 2019.08, Syntec Technology, Taiwan