Sara Vicente

Papers

Total Citations

H-Index

About

Sara Vicente is a leading researcher in computer vision and augmented reality, with a focus on 3D scene understanding and reconstruction. Her work bridges the gap between classical geometry and modern learning-based methods, enabling efficient and accurate spatial perception for AR and robotics. Her most notable contribution, "Heightfields for Efficient Scene Reconstruction for AR" (2023, 5 citations), introduces a novel approach to 3D reconstruction from posed RGB images, offering a lightweight alternative to depth-based fusion that is both real-time and learning-driven. This work is foundational for AR applications requiring rapid, on-device scene parsing. In "AirPlanes: Accurate Plane Estimation via 3D-Consistent Embeddings" (2024, 3 citations), Vicente tackles the critical task of planar surface extraction from posed images. She demonstrates that combining popular clustering techniques with 3D-consistent embeddings yields a surprisingly strong baseline, advancing state-of-the-art in geometric understanding for downstream tasks like mapping and navigation. Her research consistently emphasizes practical, deployable solutions, making her a key figure in bringing robust 3D perception to real-world systems.