Songtao Wu

About

Songtao Wu is a rising researcher in computer vision and human-centered AI, whose work focuses on the computational modeling of human motion and interactive behavior. His primary research areas include human motion prediction, interaction recognition, and graph-based deep learning. Wu’s major contribution lies in advancing Graph Convolutional Networks (GCNs) for dynamic and compositional human motion analysis. His most cited work, “Dynamic Compositional Graph Convolutional Network for Efficient Composite Human Motion Prediction” (2023, 7 citations), introduces a novel framework that efficiently predicts complex, multi-part human movements by modeling their compositional structure—a critical step for applications in intelligent surveillance and human-robot interaction. Building on this, his 2025 paper, “Learning Mutual Excitation for Hand-to-Hand and Human-to-Human Interaction Recognition” (5 citations), tackles the challenging problem of recognizing interactive actions between multiple agents. By proposing a mutual excitation mechanism within a graph convolution architecture, Wu’s work enables more accurate modeling of the spatiotemporal dependencies in hand-to-hand and human-to-human interactions. Though early in his career, Wu’s research is already shaping how machines understand and anticipate human motion, laying the groundwork for more responsive and intuitive AI systems.

Research Focus

Key Achievements

Top Papers

1. 1
   Dynamic Compositional Graph Convolutional Network for Efficient Composite Human Motion Prediction

   7 citations · 2023
2. 2
   Learning Mutual Excitation for Hand-to-Hand and Human-to-Human Interaction Recognition

   5 citations · 2025

Key Collaborators

Contact & Links