Papers
9
Total Citations
1,130
H-Index
7
About
Angela Dai is a leading researcher in 3D scene understanding, reconstruction, and geometric deep learning, with work that has profoundly shaped how machines perceive and interpret the physical world. She is best known for **BundleFusion**, a landmark system for real-time, globally consistent 3D reconstruction of large-scale scenes — a problem central to mixed reality and robotics. BundleFusion addressed the critical challenge of drift in pose estimation, enabling high-quality scanning without the costly offline processing required by prior methods; across its various publications, it has accumulated over 1,000 citations, establishing it as one of the field's cornerstone contributions. Beyond reconstruction, Dai has made significant strides in scene completion and semantic understanding. Her work on **RevealNet** tackled the challenge of inferring hidden object geometry from RGB-D scans, while **3D-SIC** introduced the novel task of semantic instance completion from incomplete scans. More recently, she has pushed toward richer holistic scene understanding through panoptic 3D reconstruction from single images and urban-scale panoptic scene completion with uncertainty modeling. Collectively, her research bridges the gap between raw 3D sensor data and structured, semantically meaningful scene representations — laying essential groundwork for intelligent robotic and augmented reality systems.
Research Focus
Key Achievements
Top Papers
- 1BundleFusion680 citations · 2017
- 2BundleFusion245 citations · 2017
- 3
- 4RevealNet: Seeing Behind Objects in RGB-D Scans64 citations · 2020
- 5PaSCo: Urban 3D Panoptic Scene Completion with Uncertainty Awareness25 citations · 2024
- 6Panoptic 3D Scene Reconstruction From a Single RGB Image8 citations · 2021
- 7RevealNet: Seeing Behind Objects in RGB-D Scans7 citations · 2019
- 83D-SIC: 3D Semantic Instance Completion for RGB-D Scans6 citations · 2019
- 9