Chunxia Xiao

Papers

Total Citations

222

H-Index

About

Chunxia Xiao is a leading researcher in 3D computer vision and geometric processing, with a primary focus on point cloud shape completion—a critical challenge for robotics, autonomous navigation, and augmented reality. Her work addresses the fundamental tension between global shape reconstruction and local detail preservation in incomplete 3D scans. Xiao’s most influential contribution, "Detail Preserved Point Cloud Completion via Separated Feature Aggregation" (2020), has garnered over 155 citations, establishing a new paradigm by decoupling global structure from local geometric details during feature learning. This approach overcame the limitations of earlier encoder-decoder methods that produced overly smooth, detail-lacking outputs. Building on this foundation, Xiao introduced the "Skeleton-Detail Transformer" (2022), which further refines completion quality by explicitly modeling the relationship between a shape’s skeletal structure and its surface details. This work has quickly accumulated 59 citations, reflecting its impact on the field. Through these innovations, Xiao has advanced the state-of-the-art in 3D completion, enabling more accurate and visually faithful reconstructions that preserve fine geometric features—a crucial step toward practical deployment in real-world 3D perception systems.