Kwanyong Park
Papers
2
Total Citations
11
H-Index
2
About
Kwanyong Park is a researcher advancing the frontiers of computer vision, with a primary focus on domain adaptation and test-time adaptation for semantic segmentation—critical technologies for autonomous driving, robotics, and medical imaging. His most influential work, "Discover, Hallucinate, and Adapt: Open Compound Domain Adaptation for Semantic Segmentation" (2021, 7 citations), tackles the challenging problem of adapting segmentation models to multiple unknown target domains simultaneously, introducing a novel framework that discovers domain-specific features, hallucinates new domains, and adapts robustly. This work addresses a key limitation in prior unsupervised domain adaptation methods, which assume a single target domain. Park further extends this line of inquiry in "Test-Time Adaptation in the Dynamic World With Compound Domain Knowledge Management" (2023, 4 citations), where he addresses the practical need for lifelong model adaptation during deployment. This work proposes a system that manages knowledge from multiple encountered domains, enabling a model to continuously adapt to novel environments without forgetting previously learned information. Park’s research is notable for its direct relevance to real-world deployment of deep recognition models, where encountering diverse, shifting conditions is the norm rather than the exception.
Research Focus
Key Achievements
Top Papers
- 1
- 2