Zhihua Tian
M.Sc. Candidate, Nankai University
I am currently a Master of Science candidate in Operations Research at Nankai University, advised by Prof. Jianda Han and Prof. Weiguang Huo. Prior to this, I received my B.E. in Computer Science from Hangzhou Dianzi University.
Biography
My research interests lie at the intersection of Computer Vision, Large Language Models (LLMs), and Healthcare. I am particularly interested in developing robust and interpretable AI systems for disease diagnosis and human-machine interaction.
Recent News Updates
- 2025-11 Accepted paper in npj Digital Medicine (IF=15.1) as Co-First Author.
- 2025-08 Accepted paper at 2025 IEEE International Conference on Cyborg and Bionic Systems (First Author).
Publications
Experience
Research Experience
Developed a patented, expert-level deep learning framework (AUC=0.87, F1=0.806) to predict Parkinson’s severity from smartphone videos. This system successfully distinguishes subtle medication (on/off) effects (73.68% acc.) at a physician-level resolution and discovers novel digital biomarkers sensitive to disease progression.
Proposed the IMU2SKE framework to accurately predict Freezing of Gait (FoG) from monocular videos. leveraged IMU-guided hierarchical contrastive learning to significantly enhance prediction accuracy from 79.5% to 83.2%.
Industry Experience
Built a voice appointment agent and a dietary comment LLM (Base: Qwen2.5-VL-7B) to improve efficiency for medical staff.
Developed a high-definition portrait generation system (Unreal Engine MetaHuman) and improved model accuracy using diffusion models (DDPM) for data generation.
Developed human body/face beauty model and algorithm, with several models and algorithms serving 2M+ photographers.
Projects Demo
Voice Appointment Agent
Developed a real-time, interruptible Voice Appointment Agent leveraging ASR, TTS, and AutoGen function calling.
Gait Analysis System
A deep learning-enabled system for accurate assessment of gait impairments in Parkinson's disease.
Awards
© 2025 Zhihua Tian