Ruoxi Shi

Papers

Total Citations

H-Index

About

Ruoxi Shi is a researcher advancing the frontier of 3D computer vision, with a primary focus on unsupervised keypoint detection and geometric deep learning. Her most recognized contribution is the development of "Skeleton Merger," an unsupervised aligned keypoint detector introduced in 2021. This work addresses a fundamental challenge in 3D analysis: detecting consistent keypoints across object instances without requiring expensive, manually annotated datasets. By leveraging the inherent skeletal structure of objects, Shi’s method enables robust alignment for applications ranging from object tracking to shape retrieval and robotics. The paper has garnered 35 citations, reflecting its growing influence among researchers seeking data-efficient solutions for 3D correspondence. Shi’s approach is notable for circumventing the ambiguity of keypoint definitions, a long-standing bottleneck in the field. Her work demonstrates a commitment to unsupervised learning paradigms, reducing reliance on labor-intensive labeling while maintaining high performance. For students and researchers exploring 3D perception, Shi’s contributions offer a compelling pathway toward scalable, annotation-free geometric understanding—a critical step for autonomous systems and interactive environments.