Avatar

Woo Jae Kim

Ph.D. Candidate

School of Computing, KAIST


CV | Google Scholar | GitHub


I am a Ph.D. student in Scalable Graphics, Vision, and Robotics (SGVR) Lab at KAIST, advised by Prof. Sung-Eui Yoon. I received my B.S. degree in Computer Science from KAIST in 2021 with minor in Electrical Engineering. My research interest lies on adversarial attack and defense in computer vision applications. I focus on investigating the vulnerability of deep neural networks via adversarial attack and designing adversarial defense techniques that can ensure robust deep model inferences under such threat.

Contact

  • wkim97 [at] kaist.ac.kr

  • Bldg E3-1, Rm 3446, 291 Daehak-ro, Yuseong-gu, Daejeon, Korea, 34141

Education

  • Ph.D in Computer Science, KAIST

    - Advisor: Sung-Eui Yoon

    Mar. 2023 - Current

  • M.S. in Computer Science, KAIST

    - Advisor: Sung-Eui Yoon

    Mar. 2021 - Feb. 2023

  • B.S. in Computer Science, KAIST

    - Minor in Electrical Engineering

    Sep. 2016 - Feb. 2021

News

  • [Jul 2023] Our work on adversarial defense is invited for oral and poster presentations at KCCV 2023.
  • [Jul 2023] One paper is accepted to ICCV 2023.
  • [Feb 2023] One paper is accepted to CVPR 2023 as a highlight paper.
  • [Feb 2023] One paper is accepted to i3D 2023.
  • [Jun 2022] One paper is accepted to ICIP 2022 as an oral paper.
  • [Mar 2022] One paper is accepted to CVPR 2022.
  • [Jan 2022] One paper is accepted to ICASSP 2022.

Publications

  • Towards Content-based Pixel Retrieval in Revisited Oxford and Paris

    Guoyuan An, Woo Jae Kim, Saelyne Yang, Rong Li, Yuchi Huo, and Sung-Eui Yoon

    ICCV 2023

    [Project] [Paper] [Datasets and Code] [Interactive Tools]

  • Feature Separation and Recalibration for Adversarial Robustness

    Woo Jae Kim, Yoonki Cho, Junsik Jung, and Sung-Eui Yoon

    CVPR 2023 Highlight paper (~2.5% acceptance rate)

    [Project] [Paper] [Code]

  • Pixel-wise Guidance for Utilizing Auxiliary Features in Monte Carlo Denoising

    Kyubeom Han, Olivia G. Odenthal, Woo Jae Kim, and Sung-Eui Yoon

    i3D 2023

    also published at Proceedings of the ACM on Computer Graphics and Interactive Techniques (PACMCGIT)

    [Project] [Paper] [Code]

  • Diverse Generative Perturbations on Attention Space for Transferable Adversarial Attacks

    Woo Jae Kim, Seunghoon Hong, and Sung-Eui Yoon

    ICIP 2022 Oral paper (~10% acceptance rate)

    [Project] [Paper] [Code]

  • Part-based Pseudo Label Refinement for Unsupervised Person Re-identification

    Yoonki Cho, Woo Jae Kim, Seunghoon Hong, and Sung-Eui Yoon

    CVPR 2022

    [Project] [Paper] [Code]

  • Deep Video Inpainting Guided by Audio-Visual Self-Supervision

    Kyuyeon Kim, Junsik Jung*, Woo Jae Kim*, and Sung-Eui Yoon (* equal contributions)

    ICASSP 2022

    [Project] [Paper] [Code]

Awards & Honors

  • Paper Award, 34th Workshop on Image Processing and Image Understanding (IPIU), 2022
    • - Developed an adversarially robust pooling operation that can replace various downsampling operations of CNN.
    - 적대적 공격에 강인한 피처 신뢰도 기반 다운샘플링. In IPIU, 2022
  • Best Teaching Assistant Award, School of Computing, KAIST, 2021
    • - Worked as a teaching assistant for CS206: Data Structure.
  • Grand Prix (1st place), Undergraduate Research Program, KAIST, 2021
    • - Developed a highly transferable black-box adversarial attack based on diverse image attention disruption.
    - Diverse Generative Perturbations on Attention Space for Transferable Adversarial Attacks. In ICIP, 2022
  • National Government Fellowship, South Korea Government, 2016 - Present

Invited Talks & Presentations

Experiences

Reviewer

  • CVPR (2023)
  • ICCV (2023)