Education

Seoul National University (SNU)
Sep. 2023 - Present, Computer Science and Engineering, Combined M.S. and Ph.D.
• Adviser: Jaesik Park

Pohang University of Science and Technology (POSTECH)
Feb. 2023 - Aug. 2023, Computer Science and Engineering, M.S.
• Adviser: Jaesik Park
Feb. 2019 - Feb. 2023, Computer Science and Engineering, B.S.
• Summa Cum Laude

Korean Minjok Leadership Academy (민족사관고등학교)
Feb. 2016 - Feb. 2019, High School

International Publications

• Fill-Up: Balancing Long-Tailed Data with Generative Models (Under Review)
  Joonghyuk Shin, Minguk Kang, Jaesik Park
  We propose a two-stage training procedure for long-tailed recognition based on the fill-up operation with recent large-scale text-to-image synthesis models. During the Stage I, the image classifier is trained on balanced dataset, obtained by filling up long-tailed data with synthetic images from textual-inverted tokens. Stage II fine-tunes the image classifier with long-tailed real dataset along with Balanced Softmax loss. Our method achieves state-of-the-art results on standard long-tailed benchmarks when trained from scratch.
  [Paper | Code | Project Page]

• StudioGAN: A Taxonomy and Benchmark of GANs for Image Synthesis (TPAMI 2023)
  Minguk Kang, Joonghyuk Shin, Jaesik Park
  While training and evaluting GAN is becoming more and more important, the current GAN research ecosystem does not provide reliable benchmarks for fair and consistent evaluation. StudioGAN is a self-contained library that provides a vast number of GAN-related functionalities as modules, reproducing more than 30 popular GANs. We present extensive benchmarks of state-of-the-art GANs, following a suggested fair evaluation protocol, and open-source all the code and models.
  [Paper | Code (3300+)]

Domestic Publications

• Using Large Scale Text-to-Image Model as a Data Source for Classification (IPIU 2023) Joonghyuk Shin, Minguk Kang, Jaesik Park
  

Honors and Awards

• Award for Outstanding Poster Presentation, IPIU (2023)
  • Awarded to paper "Using Large Scale Text-to-Image Model as a Data Source for Classification"
• Summa Cum Laude, POSTECH (2023)
• National Science and Engineering Scholarship, Korea Student Aid Foundation (2021, 2022)
• Best Graduation Project, POSTECH CSE (2022)
  • Awarded to project “Large scale generative model as a data source for vision tasks”
• Silver Award UNI-DTHON Datathon, UNI-D (Union of Korean University Student for CS) (2021)
  • Competition on classifying food images
• Global Leadership Program, POSTECH CSE (2020, 2021)
  • Scholarship for academic excellence
• Best Undergraduate Research Program, POSTECH (2020)
  • Awarded to project “Neural Point Cloud Rendering of POSTECH”
• Jigok Scholarship, POSTECH (2019, 2020)

Personal