Yuanman Li
Papers
2
Total Citations
34
H-Index
2
About
Yuanman Li is a researcher specializing in deep learning, generative modeling, and human motion analysis, with a particular focus on pedestrian trajectory prediction — a critical challenge at the intersection of computer vision and intelligent systems. His most recognized contribution is **STGlow**, a flow-based generative framework that introduces a novel Dual-Graphormer architecture to capture both the diversity of individual motion behaviors and the complexity of social interactions among pedestrians. This work addresses fundamental limitations in predicting realistic, multimodal trajectories in crowded environments, with direct applications in autonomous driving, robot navigation, and video surveillance anomaly detection. STGlow has accumulated over 32 citations since its 2023 publication, reflecting strong community interest in its innovative combination of normalizing flows with graph-transformer architectures. By leveraging dual graph attention mechanisms, Li's framework advances the state of the art in modeling spatial-temporal dependencies — a notoriously difficult problem given the unpredictable nature of human movement. Li's research sits at a vital frontier where accurate pedestrian prediction directly impacts real-world safety systems, making his contributions both technically rigorous and practically consequential for the next generation of intelligent autonomous platforms.
Research Focus
Key Achievements
Top Papers
- 1
- 2