Dahyun Kang

Hello! I am a PhD student at POSTECH under the supervision of Minsu Cho. My research lies in the areas of computer vision and deep learning, especially in the problems of learning with minimal supervisions and multi-modality learning such as vision-language problems. Name pronunciation: DAH-yon. Pronouns: Ms/She/Her.

News

May 24, 2025 I will present two workshop papers in CVPR 2025.
Feb 27, 2025 DINOv2+text is accepted to CVPR.
Feb 25, 2025 A paper on web-assisted zero-shot classification is accepted to TMLR.

Education

September 2019 - Present Pohang University of Science and Technology (POSTECH), Pohang, South Korea
Integrated MS/PhD student in Computer Vision Lab
March 2015 - August 2019 Pohang University of Science and Technology (POSTECH) , Pohang, South Korea
BS in Computer Science and Engineering
Summa Cum Laude

Experience

June 2024 - November 2024 Meta (FAIR), Paris, France
Research Scientist Intern
  • Manager: Piotr Bojanowski
July 2022 - December 2022 Meta (FAIR), London, UK
Research Scientist Intern
  • Manager: Naila Murray
September 2019 - Present Computer Vision Lab, POSTECH, Pohang, South Korea
MS/PhD Student
  • Advisor: Minsu Cho

Publications

  1. DINOv2 Meets Text: A Unified Framework for Image- and Pixel-Level Vision-Language Alignment
    Cijo Jose, Théo Moutakanni,  Dahyun Kang, Federico Baldassarre, Timothée Darcet, Hu Xu, Daniel Li, Marc Szafraniec, Michaël Ramamonjisoa, Maxime Oquab, Oriane Siméoni, Huy V. Vo, Patrick Labatut,  and Piotr Bojanowski
    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2025
    Paper
  2. Memory-Modular Classification: Learning to Generalize with Memory Replacement
    Dahyun Kang, Ahmet Iscen, Eunchan Jo, Sua Choi, Minsu Cho,  and Cordelia Schmid
    Transactions on Machine Learning Research (TMLR)
    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2025 workshop on Visual Concepts     2025
    Paper Code
  3. In Defense of Lazy Visual Grounding for Open-Vocabulary Semantic Segmentation
    Dahyun Kang,  and Minsu Cho
    European Conference on Computer Vision (ECCV), 2024
    Paper Code
  4. Contrastive Mean-Shift Learning for Generalized Category Discovery
    Sua Choi,  Dahyun Kang,  and Minsu Cho
    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024
    Paper Code
  5. Distilling Self-Supervised Vision Transformers for Weakly-Supervised Few-Shot Classification & Segmentation
    Dahyun Kang, Piotr Koniusz, Minsu Cho,  and Naila Murray
    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023
    Paper Code
  6. Few-Shot Metric Learning: Online Adaptation of Embedding for Retrieval
    Deunsol Jung,  Dahyun Kang, Suha Kwak,  and Minsu Cho
    Asian Conference on Computer Vision (ACCV), 2022
    Paper Code
  7. Integrative Few-Shot Learning for Classification and Segmentation
    Dahyun Kang,  and Minsu Cho
    IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022
    Paper Code
  8. Semi-supervised Domain Adaptation via Sample-to-Sample Self-Distillation
    Jeongbeen Yoon,  Dahyun Kang,  and Minsu Cho
    Winter Conference on Applications of Computer Vision (WACV), 2022
    Paper Code
  9. Hypercorrelation Squeeze for Few-Shot Segmentation
    Juhong Min,  Dahyun Kang,  and Minsu Cho
    International Conference on Computer Vision (ICCV), 2021
    Paper Code
  10. Relational Embedding for Few-Shot Classification
    Dahyun Kang, Heeseung Kwon, Juhong Min,  and Minsu Cho
    International Conference on Computer Vision (ICCV), 2021
    Paper Code

Honors and Awards

POSTECHIAN Fellowship Award (2023)
  • Winner
Naver Ph.D. Fellowship Award (2023)
  • Winner
BK21 Best Paper Award, POSTECH CSE (2022, 2023)
  • Excellence Award - Integrative Few-Shot Learning for Classification and Segmentation (CVPR 2022)
  • Excellence Award - Relational Embedding for Few-Shot Classification (ICCV 2021)
Professional service
  • Conference reviewer: WACV {2022-2024}, CVPR {2022-2024}, ICCV 2023, ECCV 2024, ACCV 2024, Neurips {2022-2024}
  • Journal reviewer: TPAMI, IJCV