Mengyi Liu
Papers
1
Total Citations
9
H-Index
1
About
Mengyi Liu is a computer vision researcher whose work focuses on advancing 3D scene understanding through deep learning. Her key research areas include depth estimation, surface normal prediction, and multi-stage information diffusion techniques. In her most cited work, "Multi-stage information diffusion for joint depth and surface normal estimation" (2023), Liu introduced a novel framework that simultaneously predicts depth maps and surface normals by progressively refining geometric features across multiple stages. This approach addresses the challenge of ambiguity in monocular 3D reconstruction, enabling more accurate and consistent scene geometry. With 9 citations in a short time, this paper demonstrates early impact in the field. Liu's contributions are particularly notable for integrating multi-task learning with diffusion-based architectures, offering a pathway to robust, real-world applications in robotics, autonomous navigation, and augmented reality. Her work bridges the gap between geometric reasoning and modern neural network design, making her a promising voice in 3D vision research.
Research Focus
Key Achievements
Top Papers
- 1