Haojie Zhang

Papers

Total Citations

H-Index

About

Haojie Zhang is an emerging researcher at the forefront of computer vision and foundation model adaptation, with a particular focus on image segmentation and robust generalization under real-world distribution shifts. His most recognized work addresses one of the field's pressing challenges: adapting large-scale segmentation foundation models, such as Segment-Anything (SAM), to perform reliably when deployed in conditions that differ from their training environments. By leveraging weakly supervised adaptation strategies, Zhang's research offers practical pathways to extend the utility of powerful vision models without requiring costly, exhaustive annotation pipelines — a contribution with significant implications for medical imaging, autonomous driving, and remote sensing applications where labeled data is scarce. With his 2024 paper already accumulating 28 citations in a short period, Zhang has demonstrated an ability to identify timely and impactful research questions that resonate with the broader computer vision community. His work sits at the productive intersection of prompt engineering, transfer learning, and semi-supervised learning — areas experiencing rapid growth following the rise of large language and vision models. For students and researchers navigating the challenge of deploying foundation models in specialized domains, Zhang's contributions represent both an insightful analysis of current limitations and a promising direction for more adaptable, generalizable vision systems.