Shiu-Hong Kao is a first-year PhD student in Computer Science in the CVML Group at National University of Singapore (NUS), advised by Prof. Angela Yao. His research focuses on video understanding, reasoning segmentation, and LLM/VLM reasoning.
Shiu-Hong Kao is a first-year PhD student in Computer Science in the CVML Group at National University of Singapore (NUS), advised by Prof. Angela Yao. His research sits at the intersection of deep learning and computer vision, with a focus on video understanding, reasoning segmentation, and LLM/VLM reasoning.
He received the M.Phil. in Computer Science and Engineering from the Hong Kong University of Science and Technology (HKUST) under the supervision of Prof. Chi-Keung Tang, collaborating closely with Prof. Yu-Wing Tai from Dartmouth College. He also conducted research in 3D reconstruction and interned in the Media Computing Group at Microsoft Research Asia, working with Dr. Xiao Li and Dr. Jinglu Wang.
Before graduate school, he earned a Bachelor of Science degree from HKUST, double-majoring in Data Science and Technology and Computer Science. He completed his undergraduate thesis in Prof. Shueng-Han Gary Chan's lab, in cooperation with Dr. Jierun Chen.
About his name: He goes by Daniel in daily life. His Chinese name, Shiu-Hong, is typically used only in official documents and publications. In mainland China, he is sometimes referred to as Xuhong, based on a different homophonic transliteration system.
International Conference on Learning Representations (ICLR), 2026
We propose CoT-RVS to extract the temporal-semantic correlation in videos with chain of thoughts and achieve the state-of-the-art performance for reasoning video segmentation.
ICLR 2026 Workshop on AI with Recursive Self-Improvement (RSI 2026)
*Equal contribution. We propose CoT-Seg, a modular reasoning segmentation framework that significantly improves the quality of segmentation masks with chain-of-thought reasoning and a self-correction loop.
International Conference on Computer Vision (ICCV), 2025
We propose StreamGS, an online generalizable 3DGS reconstruction method for unposed image streams, which progressively transforms image streams to 3D Gaussian streams by predicting and aggregating per-frame Gaussians.
Conducted research in 3D computer vision and contributed to research publications in the community.
Designed and developed algorithms across image processing, OCR, data parsing, and deep learning. Report