My name is Wenhao Zheng (郑文浩). I am currently a PhD student at The University of North Carolina at Chapel Hill majoring in Computer Science, supervised by Prof. Huaxiu Yao. Before that, I graduated as a master’s student in the Computer Science department at Zhejiang University, supervised by Prof. Jian Wu in 2024 and graduated with a B.E. degree in Computer Science and Technology from Chu Kochen Honors College at Zhejiang University in 2021.

My current research interest primarily includes Computer Vision (CV), Natural Language Processing (NLP) based on LLMs and Multi-Modal Learning (MML) involving Vision and Language modalities, especially for Healthcare (AI4H). Specifically, I have published several papers on the following topics:

  • Tabular data prediction and text classification based on LLM[NLP].
  • Weakly-supervised instance segmentation on medical images[CV, AI4H].
  • Object detection with multiple modalities for healthcare[CV, MML, AI4H].
  • Image classification with ordinal regression and generation for healthcare[CV, AI4H].

In addition, I am also interested in the following topics:

  • High-performance AI systems, including AI chips, operators in AI framework and performance optimization in distributed training.
  • Reinforcement learning, especially for the application in LLM and games, such as StarCraft II.
  • Physics, physical simulation and physics-guided learning.
  • Any interesting technical topics in computer science, physics and finance I am not familiar with now.

In my spare time, I also enjoy writing some helpful and interesting tools for research and daily life. You can find them on my GitHub.

🔥 News

  • 2024.02: 🎉 One paper was released on arXiv.
  • 2023.12: 🎉 One paper was accepted by JBHI.
  • 2023.10: 🎉 One paper was accepted by TMI.
  • 2023.01: 🔥 A helpful library with a collection of tools wrapped on pytorch-lightning was released.
  • 2022.05: 🎉 One paper was accepted by TCBB.
  • 2022.03: 🎉 One paper was accepted by TMI.

📝 Publications

($\dagger$: Equal contribution; $^*$: Corresponding author(s))

JBHI
sym

It is the first work to propose a bespoke approach, called PAL, for instance segmentation of convex objects, leveraging the convex shapes of biomedical targets to generate the profile (i.e., masks) with bounding boxes. PAL outperforms other box-supervised models (e.g., BoxInst), and achieves comparable performance with mask-supervised models including Mask R-CNN and Cascade Mask R-CNN, on convex object instance segmentation tasks. In addition, we show that PAL demonstrates promising performance even for non-convex objects in practical applications.

Polygonal Approximation Learning for Convex Object Segmentation in Biomedical Images with Bounding Box Supervision[CV, AI4H], Wenhao Zheng, Jintai Chen, Kai Zhang, Jiahuan Yan, Jinhong Wang, Yi Cheng, Bang Du, Danny Z. Chen, Honghao Gao$^*$, Jian Wu, Hongxia Xu$^*$, IEEE Journal of Biomedical and Health Informatics, 2023.

TMI
sym

We propose a novel task-decomposing and cell-comparing framework for cervical lesion cell detection. The original detection task is decomposed into two detection subtasks first and a decompose-and-integrate head is introduced to model those tasks encouraging the network to focus on specific cell structures. In addition, cell comparison is performed by introducing a dynamic comparing module and an instance contrastive loss for both normal-abnormal and abnormal-abnormal cells comparison, enhancing the ability of the model to distinguish different kinds of lesion cells.

A Task Decomposing and Cell Comparing Method for Cervical Lesion Cell Detection[CV, MML, AI4H], Tingting Chen$^\dagger$, Wenhao Zheng$^\dagger$, Haochao Ying, Xiangyu Tan, Kexin Li, Xiaoping Li, Danny Z. Chen, Jian Wu$^*$, IEEE Transactions on Medical Imaging, 2022.

💻 Projects

sym

A generic project template library with helpful and powerful tools based on pytorch-lightning for project-template project. It is designed to help AI researchers build their models quickly and easily without repeating boilerplate engineering codes, like managing and comparing config files between experiment runs, searching hyperparameters automatically and implementing cross-validation, performance profiler and loggers.

📈 More statistics

🎖 Honors and Awards

  • 2022.11 Zhijun He Scholarship, Zhejiang University
  • 2022.10 First Class Academic Excellence Scholarship, Zhejiang University
  • 2019.08 1$^{st}$ Prize in The 12$^{th}$ World Robotics Sailing Championship(WRSC)

📖 Educations

  • 2021.09 - now, M.E. in Software Engineering, College of Computer Science and Technology, Zhejiang University.
  • 2017.09 - 2021.06, B.E. in Computer Science and Technology, Chu Kochen Honors College, Zhejiang University.