Xiaofei Shao

Papers

Total Citations

H-Index

About

Xiaofei Shao is a researcher whose work lies at the intersection of computer vision and geometric deep learning, with a primary focus on 3D scene understanding and depth completion. Her most notable contribution is the development of GraphCSPN, a geometry-aware depth completion framework that leverages dynamic graph convolutional networks (GCNs) to refine sparse depth maps into dense, accurate predictions. This work, published in 2022 and already garnering 37 citations, addresses a critical challenge in autonomous driving and robotics: reconstructing fine-grained 3D geometry from incomplete sensor data. By integrating spatial relationships through graph-based reasoning, Shao’s approach achieves state-of-the-art performance on benchmark datasets, demonstrating how structured geometric priors can enhance learning-based depth estimation. Her research bridges the gap between traditional geometric methods and modern deep learning, offering practical solutions for real-world perception systems. With her innovative use of dynamic GCNs, Shao is shaping the future of depth-aware vision, making her work essential reading for students and researchers exploring robust 3D reconstruction techniques.