Dongjun Hwang

[linkedin] [google scholar] [github]

(Updated in 2025.01) Curriculum vitae can download HERE.

image

About Me

I am currently a Ph.D. Student in the Department of Computer Science and Engineering at Sogang University, supervised by Prof. Junsuk Choe. Prior to this, I obtained my M.S. Degree in the Department of Artificial Intelligence at Sogang University advised by the same supervisor. Previously, I received a B.S. Degree at the same university.

Research Interests: I research how to make open-vocabulary dense prediction models, such as segmentation and object detection, more scalable. My focus is on enabling visual grounding models to perform lifelong learning. In addition, I explore how to scale dense prediction models using imperfect datasets, such as self-supervised, semi-supervised, weakly supervised, and few-shot learning.

My recent primary research question revolves around improving performance in unseen domains through continual learning.

Contact: djhwang(at)sogang(dot)ac(dot)kr


News

  • 2025.01: 1 paper accepted in Pattern Recognition Letters. (SJR Q1; IF: 3.90)
  • 2024.10: I have released the preprint “Overcoming Domain Limitations in Open-vocabulary Segmentation” on arXiv. Please take a look!
  • 2024.09: I have received a grant from the NRF Fellowship for PhD students.
  • 2024.08: I have been accepted as a Visiting Student at CMU through a program organized by IITP.
  • 2023.03: 1 paper accepted in Pattern Recognition Letters. (SJR Q1; IF: 4.75)

Publications

* Equal Contribution. † Corresponding Author.


  • A New Benchmark for Few-Shot Class-Incremental Learning: Redefining the Upper Bound

    Shiwon Kim*, Dongjun Hwang*†, Sungwon Woo *, Rita Singh†. [Preprint]

    • arXiv preprint: 2503.10003


  • Overcoming Domain Limitations in Open-vocabulary Segmentation

    Dongjun Hwang, Seong Joon Oh, Junsuk Choe†. [Preprint] [Code]

    • arXiv preprint: 2410.11536


  • Small Object Matters in Weakly Supervised Object Localization

    Dongjun Hwang, Seong Joon Oh, Junsuk Choe†.

    • Submitted to Neurocomputing. (Under Revision)


  • Curriculum Learning with Class-label Composition for Weakly Supervised Semantic Segmentation

    Dongjun Hwang, Hyoseo Kim, Doyeol Baek, Hyunbin Kim, Inhye Kye, Junsuk Choe†. [Journal]

    • Pattern Recognition Letters, 2025. (SJR Q1; IF: 3.90)


  • Entropy Regularization for Weakly Supervised Object Localization

    Dongjun Hwang, Jung-Woo Ha, Hyunjung Shim, Junsuk Choe†. [Journal]

    • Pattern Recognition Letters, 2023. (SJR Q1; IF: 4.75)

Education

  • Ph.D. in Department of Computer Science and Engineering, Sogang University, Mar. 2024 – Present
  • M.S. in Department of Artificial Intelligence, Sogang University, Mar. 2022 – Feb. 2024
    • Advisor: Junsuk Choe
    • Thesis: Curriculum Learning with Class-label Composition for Weakly Supervised Semantic Segmentation
  • B.S. in Department of Computer Science and Engineering, Sogang University, Mar. 2018 – Feb. 2022

Experiences

  • Visiting Student at Carnegie Mellon University, IITP AI Intensive Education Program, Aug. 2024 - Feb. 2025
    • Project 1 (Lecture Code: 11-785): Invisible Watermarks: Attacks and Robustness [Report]
    • Project 2 (Lecture Code: 11-775): TBD

Awards & Scholarship