Yongyi Su

About

Yongyi Su is a rising researcher in computer vision and machine learning, with a focus on foundation models for image segmentation and their robustness under distribution shifts. His most-cited work, "Improving the Generalization of Segmentation Foundation Model under Distribution Shift via Weakly Supervised Adaptation" (2024, 28 citations), addresses a critical limitation of models like Segment-Anything (SAM)—their vulnerability to real-world domain shifts. Su introduces a weakly supervised adaptation framework that enhances SAM’s zero-shot generalization without requiring dense annotations, making segmentation more reliable in diverse, unconstrained environments. This contribution bridges the gap between foundation model flexibility and practical deployment, particularly in medical imaging, autonomous driving, and remote sensing. By leveraging minimal supervision, Su’s approach reduces annotation costs while preserving performance, a key step toward scalable, trustworthy AI. His work has quickly gained attention for its pragmatic yet impactful methodology, positioning him as a promising voice in the next generation of vision foundation model research.

Research Focus

Key Achievements

Top Papers

1. 1
   Improving the Generalization of Segmentation Foundation Model under Distribution Shift via Weakly Supervised Adaptation

   28 citations · 2024

Key Collaborators

Contact & Links