The University of Hong Kong
Master of Science in Artificial Intelligence
Core coursework: Deep Learning, Natural Language Processing, Computer Vision, and AI Ethics.
About Me
I am Waikit Xiu, an M.Sc. student in Artificial Intelligence at The University of Hong Kong. I received my B.Eng. in Traffic Engineering from Sun Yat-sen University. My research interests lie in multimodal visual understanding and embodied world models, with an emphasis on large-scale video understanding in complex dynamic environments and efficient world-model inference.
My work centers on video understanding in complex dynamic environments. I study how models can capture spatio-temporal relationships and environmental change from continuous visual streams, while supporting longer temporal horizons and more complex reasoning under practical compute constraints. My current interests include multimodal spatio-temporal understanding, fine-grained vision-language modeling, and efficient world models.
I am seeking Ph.D. opportunities for Fall 2027 and welcome conversations with faculty and groups working on these problems.
News
🎉 Our explainable semantic sensing paper was accepted to IEEE Sensors Journal — an exciting milestone.
🚀 New chapter unlocked — joining HKUST (Guangzhou) as RA in the Transportation Thrust. No cap, I'm hyped. LFG!
🎉 NanoVerse-TSR accepted to IEEE Sensors Journal! All that grind finally paid off — let's keep going!
✨ Sony internship done. Learned a great deal, met great people — genuinely one of the best summers I've had.
🎉🎉 LASAR was accepted to CVPR 2026! Looking forward to Nashville! 🔥
Education
Master of Science in Artificial Intelligence
Core coursework: Deep Learning, Natural Language Processing, Computer Vision, and AI Ethics.
Bachelor of Traffic Engineering
GPA: 3.6/4.0. Relevant coursework: Computer Vision, Image Processing, Deep Learning, Data Structures and Algorithms, and Machine Learning.
Selected Publications & Patent
NanoVerse-TSR: Contrastive Learning-Driven Traffic Sign Recognition.
Explainable Semantic Sensing for Traffic Safety Risk Detection: A Vision-Language Approach.
自动驾驶场景中周围车辆的轨迹预测方法、设备及介质
* Equal contribution.
Manuscripts in Preparation
PredErase: Training-Free Object-and-Effect Removal with Predictive Latent Guidance.
Traffic-MLLM: Curiosity-Regularized Supervised Learning for Traffic Scenario Case-Based Reasoning.
Work Experience
Research Assistant, Transportation Thrust
Research Intern, MLLM & Generative AI Team
Embodied AI Intern, Vision & Perception Group
Affiliations
Reviewer, Frontiers in Plant Science (JCR Q1).
Reviewer, IEEE ICASSP 2025.
Reviewer, IJCNN 2026.
Projects
Principal Researcher
Developed a multi-modal state prediction system integrating LiDAR point clouds and camera images to generate bird's-eye view representations for real-time scene understanding and dynamic agent behavior prediction. The project was recognized as a national-level innovation project with excellent final results and produced an invention patent.
Lead Developer
Led the development of an autonomous smart vehicle system using ROS, covering visual perception, SLAM-based navigation, and tracking and control. The project received university-level recognition and won first prize in the South China regional competition and second prize in the national final of the 18th National College Student Smart Car Competition.
Technical Toolkit
Core programming, AI development, robotics, and system operations skills.
Research & automation
Systems development
Deep learning
AI-assisted building
Robotics middleware
Shell & environments
Container workflows
Operations & automation
Version control
Awards & Honors