Guangyao Zhou
Papers
2
Total Citations
34
H-Index
2
About
Guangyao Zhou is a researcher at the forefront of robotics and computer vision, with a focus on enabling machines to perceive and interact with the physical world more intelligently. His work spans two critical areas: **few-shot visual imitation learning** and **probabilistic 3D scene understanding**. Zhou’s most notable contribution is **RoboTAP**, a novel framework for tracking arbitrary points in video to allow robots to learn new behaviors from just a single human demonstration. This work, which has already garnered **31 citations** since its 2024 publication, directly tackles the long-standing challenge of making robots adaptable in unstructured, real-world environments without task-specific engineering. In parallel, Zhou has advanced **6D pose estimation** through his work on **3D Neural Embedding Likelihood (3DNEL)**, which introduces probabilistic inverse graphics to robustly infer object pose from 2D images—a crucial capability for reliable robotic manipulation. By bridging the gap between data-efficient learning and robust 3D perception, Zhou’s research is paving the way for robots that can be quickly and intuitively taught new skills, moving beyond the confines of specialized labs and factories.
Research Focus
Key Achievements
Top Papers
- 1RoboTAP: Tracking Arbitrary Points for Few-Shot Visual Imitation31 citations · 2024
- 2