Ph.D. Student · Computer Science
I’m Hui Zhu, a researcher at Wayne State University working across computer vision, multimodal learning, and trustworthy AI.
Visual representation learning across modalities and domains.
My research asks how models can learn from different kinds of visual information—and still remain reliable when the data, domain, or task changes.
I am a Computer Science Ph.D. student at Wayne State University, advised by Prof. Dongxiao Zhu in the Trustworthy AI Lab. My work spans image segmentation, vision-language learning, foundation models, and AI evaluation.
I enjoy turning difficult research questions into reproducible systems: building data pipelines, designing controlled experiments, and translating results across disciplines.
Selected research
From modalities to geographies to model access, my work focuses on learning under real-world constraints.
A pixel-grounded vision-language model that combines conversational guidance, segmentation, and relative depth estimation for accessible pedestrian navigation.
Read the paper ↗Fine-tuning Segment Anything with sparse and dense visual prompting for mobility infrastructure in aerial imagery.
View on arXiv ↗A practical calibration framework for more reliable membership inference attacks on large language models.
View on arXiv ↗Publications
Rafi Ibn Sultan, Hui Zhu, Xiangyu Zhou, Chengyin Li, et al.
Saleh Zare Zade, Yao Qiang, Xiangyu Zhou, Hui Zhu, et al.
Rafi Ibn Sultan, Hui Zhu, Chengyin Li, Dongxiao Zhu.
Karlin Hilai, Daniil Grubich, Marcus Akrawi, Hui Zhu, et al.
Karlin Hilai, Daniil Grubich, Marcus Akrawi, Hui Zhu, et al.
Chengyin Li, Hui Zhu, Rafi Ibn Sultan, et al.
Rafi Ibn Sultan, Chengyin Li, Hui Zhu, et al.
Beyond research
I care about the work around the model: clear collaboration, reproducible experiments, and helping technical teams move from an idea to a result.
Wayne State University · CSC 4996 Capstone Projects
Mentoring student teams through product scoping, implementation, milestones, and technical presentation.Trustworthy AI Lab · Wayne State University
Developing and evaluating deep learning methods in interdisciplinary research teams.Wang Laboratory · Dalian University of Technology
Deep learning for signal analysis, algorithm reproduction, and experimental validation.Toolkit
Python · C/C++ · Bash · SQL
PyTorch · MONAI · Transformers · scikit-learn
U-Net · ResNet · Vision Transformers · SAM
Linux · Docker · CUDA · HPC · Git
Let’s connect
If you’re working on computer vision, multimodal learning, or trustworthy AI, I’d be glad to hear from you.
hui@wayne.edu