Papers
19
Total Citations
827
H-Index
10
About
Siyu Tang is a prominent computer vision and machine learning researcher whose work spans gaze estimation, 3D human body understanding, human-scene interaction, and robotic grasping. Based at ETH Zurich, Tang has made significant strides in developing datasets and methods that push the boundaries of how machines perceive and model human behavior in complex, real-world environments. Among her most influential contributions is the ETH-XGaze dataset (2020, 297 citations), a large-scale benchmark for gaze estimation under extreme head pose and gaze variation that has become a key reference for the research community. Her work on "Grasping Field" (2020, 190 citations) introduced novel implicit representations for synthesizing realistic human grasps, addressing the longstanding challenge of modeling the hand's many degrees of freedom. Further extending her reach into embodied AI, Tang has tackled 4D human body capture in 3D scenes, egocentric body estimation via head-mounted devices, and stochastic whole-body grasping through the SAGA framework. Her contributions to 3D point cloud segmentation and fine-grained action parsing round out a research portfolio that bridges perception, synthesis, and interaction — making her work essential reading for anyone working at the intersection of computer vision, robotics, and human-centered AI.
Research Focus
Key Achievements
Top Papers
- 1
- 2Grasping Field: Learning Implicit Representations for Human Grasps190 citations · 2020
- 3Learning Motion Priors for 4D Human Body Capture in 3D Scenes83 citations · 2021
- 4
- 5SAGA: Stochastic Whole-Body Grasping with Contact54 citations · 2022
- 6Local Temporal Bilinear Pooling for Fine-Grained Action Parsing25 citations · 2019
- 7Interactive Object Segmentation in 3D Point Clouds24 citations · 2023
- 8
- 93D Segmentation of Humans in Point Clouds with Synthetic Data19 citations · 2023
- 10Grasping Field: Learning Implicit Representations for Human Grasps12 citations · 2020