CV Page

🎯 Research Focus

My core research agenda centers on human-centric artificial intelligence, with specialized expertise in:

🕴️ Human Pose Estimation: 3D Reconstruction Occlusion Handling MultiModel Methods
💃 Motion Synthesis & Generation: Interactive Text-Driven Control Long-Sequence Generation
👟 Video Motion Capture System: Multi-View High Accuracy Biomechanical Analysis

⚙️ Technical Arsenal

Meta Human: Stable Diffusion DiT VAE PPO Controllable
3D Vision: Mesh Motion Synthesis 3D Pose Estimation
Motion Capture: Multi-View Marker-less Biomechanics Analysis
Multimodal: Image Text Motion
Optimization: KV Cache LoRA Noise Strategy CFG

🚀 Key Projects

💃 Self-Forcing Autoregressive Diffusion for Real-Time Text-Driven Motion Generation

AutoRegress Diffusion DiT Reinforcement Learning (PPO) Real-time Control

✅ SOTA Performance: Achieved 300+ FPS with KV Cache acceleration
✅ Enabled continuous long-sequence generation via novel self-forcing paradigm
✅ Balanced quality/speed via noise scheduling & annealing strategies

real-time video demo

🌐 MetaPose: Multimodal Enhancement and Transformation Alignment 3D Pose Restruction

Vision-Text-Spatial Fusion Cross-Modal Alignment Occlusion Robustness

✅ IEEE TMM (Q1 Under Review)
✅ SOTA Accuracy: Outperformed video-based methods using single-frame input
✅ Novel Framework: A unified distribution space for anatomical/textual/visual features

video demo in the wild

👟 Multi-View Video Capture for Biomechanics Analysis

ViT Temporal Processing Triangulation

✅ EAAI (Q1, Major Revision)
✅Journal of Biomechanics (Top Journal of Orthopedics, Major Revision)
✅ High Accuracy: achieve the requirements of clinical biomechanical analysis.It is more accurate than the online markerless motion capture systems available on the market.

inference visualization

🤖 Provably Safe Vision-Based Teleoperation for Dual-Arm Surgical Robots

Surgical Robot Gesture Control Teleoperation

✅ IEEE ROBIO 2025

Demonstration

⚡ Connection-Aware Graph Convolution Networks

GCN Anatomical/Kinematic Modeling Multi-Level Aggregation

✅ HCIS Journal (JCR Q1, Under Review)
✅ 40% Faster than video-based approaches while maintaining SOTA accuracy

🩺 Reinforce Learning Enhanced CTA for Noninvasive Prediction

PPO CTA Medical Image Analysis

✅ Liver International (JCR Q1)
✅ High accuracy rate of identifying the ROI area of extremely severe lesions

CAM visualization

🏆 Recognitions

National Encouragement Scholarship 2020-2022 (Top 10%)
4 SCI Q1 Papers (First-Author, 3 Under Review Now)
GPA 3.5+ @ Project 985 universitiy
GPA Top 10% @ Project 211 university

📜 Publications

Accept: Liver International (Q1) ROBIO 2025🥰
Major Revision: EAAI (Q1), Journal of Biomechanics (Q2)，
Under Review: HCIS (Q1), TMM (Q1), AAAI 2026 (CCFA)
Prepare to Submit: CVPR 2026 (CCFA)