Xinyan Ma

About

Xinyan Ma is a researcher whose work lies at the intersection of computer vision and human-robot interaction, with a particular focus on human action prediction. Her key contribution, "Scene-Perception Graph Convolutional Networks for Human Action Prediction," introduces a novel framework that goes beyond traditional skeleton-based methods. While most approaches rely solely on skeletal data using recurrent neural networks, Ma’s work incorporates scene context—objects and their interactions with humans—into a graph convolutional network. This innovation allows for more accurate and context-aware predictions of human actions, a critical capability for intelligent robots and machine vision systems. Although early in its citation trajectory with 2 citations, the paper represents a forward-looking integration of scene perception into action prediction, addressing a key limitation of prior work. Ma’s research signals a shift toward richer, more holistic models that consider both human pose and environmental cues, offering a promising direction for advancing autonomous systems that can anticipate and respond to human behavior in dynamic settings.

Research Focus

Key Achievements

Top Papers

1. 1
   Scene-Perception Graph Convolutional Networks for Human Action Prediction

   2 citations · 2021

Key Collaborators

Contact & Links