Siyu Hong
Papers
1
Total Citations
9
H-Index
1
About
Siyu Hong is a researcher advancing the frontiers of computer vision and geometric deep learning, with a primary focus on 3D scene understanding from monocular imagery. Their most cited work, "Multi-stage information diffusion for joint depth and surface normal estimation" (2023), introduces a novel framework that simultaneously predicts dense depth maps and surface normals by progressively propagating geometric cues across spatial scales. This approach addresses a fundamental challenge in reconstructing 3D structure from a single image, achieving state-of-the-art accuracy on benchmark datasets. Hong’s contributions are distinguished by their elegant integration of diffusion-based reasoning with multi-task learning, enabling robust performance even in ambiguous or textureless regions. With 9 citations in under two years, this paper has quickly become a reference point for researchers working on joint estimation problems. Hong’s work is particularly notable for its practical implications in robotics, autonomous navigation, and augmented reality, where reliable 3D perception is critical. By bridging the gap between depth and normal estimation, Siyu Hong is helping to build the foundational tools for machines to see and understand the world in three dimensions.
Research Focus
Key Achievements
Top Papers
- 1