Sungwon Woo

Sungwon Woo

AI Researcher in Computer Vision
M.S. in Artificial Intelligence (Sogang University)

about me

I am an alum of Sogang University, where I earned a Master’s degree in Artificial Intelligence with a specialization in Computer Vision in 2025, and a Bachelor’s degree in Computer Science and Engineering at 2023.

My research lies at the intersection of computer vision and deep learning– with a special interest on building intelligent visual systems that are beyond supervised learning and data memory efficient. My primary research interests include continual learning, few‑shot learning, self(semi)‑supervised learning, while I also have experience in image retrieval, and lightweighting.

news

  • Oct 2025: I will attend ICCV 2025 at Honolulu, Hawaii.
  • Aug 2025: Graduated from Sogang University with a Master’s degree in Artificial Intelligence.
  • Jul 2025: One paper is accepted as a poster at ICCV 2025 workshop on Continual Learning in Computer Vision.
  • Nov 2024: One paper is accepted as a poster in WACV 2025.
  • Aug 2024: I have been granted as a visiting scholar at CMU hosted by IITP.

selected publications

  • Neural Collapse-Driven, Uncertainty-Aware Framework for Few-Shot Class-Incremental Learning
    Neural Collapse-Driven, Uncertainty-Aware Framework for Few-Shot Class-Incremental Learning
    Sungwon Woo
    M.S. Thesis, Sogang University, 2025.
    #continual learning #few-shot learning #neural collapse

    📄 Paper 🖥️ Slides

  • Does Prior Data Matter? Exploring Joint Training in the Context of Few-Shot Class-Incremental Learning
    Does Prior Data Matter? Exploring Joint Training in the Context of Few-Shot Class-Incremental Learning
    Shiwon Kim*, Dongjun Hwang*, Sungwon Woo*, Rita Singh+ (co-first author)
    Workshop on Continual Learning in Computer Vision (CLVision) at ICCV 2025
    #continual learning #few-shot learning #imbalanced learning

    📄 Paper 🎥 Video

  • Relational Self-supervised Distillation with Compact Descriptors for Image Copy Detection
    Relational Self-supervised Distillation with Compact Descriptors for Image Copy Detection
    Juntae Kim*, Sungwon Woo*, Jongho Nang+ (co-first author)"
    IEEE/CVF Winter Conference on Applications of Computer Vision 2025 (WACV25)
    #visual copy detection #knowledge distillation #image retrieval

    📄 Paper 🎥 Video

preprints

  • Traning-free Uncertainty-guided Logit Adjustment for Few-shot Class-Incremental Learning
    Traning-free Uncertainty-guided Logit Adjustment for Few-shot Class-Incremental Learning
    Under Double-blind Review. Submitted to Top-tier AI Conference.
    #continual learning #few-shot learning #uncertainty sampling

selected projects

  • AI on the Edge with Robotics
    AI on the Edge with Robotics
    Developed an autonomous targeting AI Cannon system based on NVIDIA Jetson Nano
    #robotics #object detection #object tracking

    🎥 Video

  • Invisible Watermarks: Attacks and Robustness
    Invisible Watermarks: Attacks and Robustness
    Improving watermark robustness via cascaded image- and latent-space techniques, and enhancing attacks with a custom remover network
    #invisible watermark #adversarial attack #generative ai

    📄 Paper 🎥 Video

  • Generating Yua Han’s SNS Images and Reels
    Generating Yua Han’s SNS Images and Reels
    Created virtual human Yua Han's Instagram images and reels video with Dreambooth and Face Reenactment
    #generative ai #face reenactment

    🎥 Video

  • Semi-supervised learning Knowledge Distillation with feature Augmentation (KDA)
    Semi-supervised learning Knowledge Distillation with feature Augmentation (KDA)
    Proposed a novel approach that integrates knowledge distillation with a top-k strategy to enhance classification of semi-supervised learning
    #semi-supervised learning #knowledge distillation

    📄 Paper

  • Autoencoder-based CNN-LSTM drowsy driving detection model and safe driving system using YOLOv7(5) road front object recognition
    Autoencoder-based CNN-LSTM drowsy driving detection model and safe driving system using YOLOv7(5) road front object recognition
    Development of Safety Functions Based on Driving Environment and Driver State Recognition
    #object detection #object tracking

    📄 Paper 🎥 Video 🎥 Video

Education

  • M.S. in Department of Artificial Intelligence, Sogang University, Mar.2023 - Aug.2025
    • Thesis: Neural Collapse-Driven, Uncertainty-Aware Framework for Few-Shot Class-Incremental Learning
  • B.S. in Department of Computer Science and Engineering, Sogang University, Mar.2017 - Feb.2023

Experience

  • Visiting Student at Carnegie Mellon Univeristy, Aug.2024 - Feb.2025
    • Sponsored by Institute of Information & Communications Technology Planning & Evaluation (IITP)
    • Achieved 2nd place (2/34) in the Best Student Award
  • Student Researcher at Smilegate, June.2024 - Jul.2024

Awards & Scholarship

  • Carnegie Mellon University AI Intensive Training Program Fellowship, IITP Aug.2024 - Feb.2025
  • Smilegate AI Major (DHE) Scholarship, Smilegate, Mar.2023 - Aug.2025
  • 5th place on Video Similarity Challenge, CVPRW2023, VCDW by META Jan.2023 - Mar.2023
  • 4th place on Food Classification Challenge, KT GenieLabs Aug.2022 - Sep.2022