Hyesong Choi

I am a Ph.D. student at the Computer Vision Lab, Ewha W. University, advised by Prof. Dongbo Min.
I am also a research intern at the Backbone Research Group, Naver AI Lab, advised by Dongyoon Han.

• Machine Learning Researcher, Specializing in Training Efficiency

My research centers on one of the most fundamental challenges in machine learning: efficiency.

I focus on reducing training time, improving representation quality, and scaling models effectively across key ML paradigms, including large model pre-training, transfer learning to recognition tasks, and reinforcement learning.

- Generative Pre-training for Efficient Representation Learning
In my Dissertation and [P8, P9], I proposed generative pre-training frameworks that integrate Denoising Diffusion Models (DDMs) with large-scale model training. These methods significantly improved representation quality, achieving up to 3.4% accuracy gains across multiple downstream tasks.

- Corruption-based Learning for Faster Convergence
In my Dissertation and [C6, C7], I developed masking techniques that reduced training time by up to 50%, while maintaining strong representation performance, proposing an efficient and scalable ML approach to large-model pre-training.

- Sample-Efficient Reinforcement Learning for Robotics Simulation

My earlier works [C4, C5] and [C8] explored how computer vision can be integrated with reinforcement learning to improve sample efficiency and representation quality in robotics-related simulation environments as DeepMind Control Suite, Atari Games, and DrawerWorld. This research addresses a key problem in ML: enabling agents to perceive and act effectively from high-dimensional visual inputs.

Publications (International)

• Conference

• Journal

• Preprint

Teaching

[2025 Spring] Computational Thinking and Programming, CC department, Ewha W. University
(Main Lecturer)

[2024 Fall] Contents Database, CC department, Ewha W. University
(Main Lecturer)

[2024 Spring] Data Innovation and AI Technology, EB department, Seoul Media Institute of Technology  - Lectured in English
(Main Lecturer)

Education

• M.S./Ph.D. [2020.3 ~ Current] (Advisor: Dongbo Min)
  Computer Science
  (Distinguished Research Scholarship, Awarded across the entire Graduate School)
  Ewha W. University (EWU)

• B.S. [2016.3 ~ 2020.2]
  Contents Convergence, Computer Science
  (Graduated Summa Cum Laude, rank: 1/n)
  Ewha W. University (EWU)

Professional Experience

• 2025.3 - Current: Research Intern, Backbone Research Group, Naver AI Lab (Advisor: Dongyoon Han)

• 2024.9 - Current: Instructor, Ewha W. University (EWU)
  (Teaching: Contents Database, Computational Thinking and Programming)

• 2024.7 - 2024.9: Visiting Researcher, University of British Columbia (UBC) (Advisor: Kwang Moo Yi, Research topic: Generative Pre-Training of Large Vision Models)

• 2024.3 - 2025.2: Adjunct Professor, Seoul Media Institute of Technology (SMIT)
  (Teaching: Data Innovation and AI Technology)

• 2021.4 - 2022.3: Research Assistant, Hyundai Motors Group
  (Project topic: AI-based Automatic Water Inflow Detection Technology to Enhance Vehicle Development Efficiency)

• 2020.4 - 2021.12: Research Assistant, Institute for Information communication Technology Planning and Evaluation (IITP)
  (Project topic: Development of Artificial Intelligence Technology that Continuously Self-improves According to Changing Situations in the Real World)

• 2020.3 - 2021.2: Research Assistant, National Research Foundation of Korea
  (Project topic: Multi-task Learning Model Study for Complementary Scene Understanding based on Depth Estimation and Object Detection/Segmentation)

• 2018.8 - 2018.12: Research Intern, Action Power
  (Research topic: Development of Network Architectures and Augmentation Techniques for Audio Data)

Honor and Award

• Boeing-Ewha Scholarship, Received Study Incentives (total funding: 4 million KRW), 2025

• Distinguished Research Scholarship, Awarded across the entire Graduate School at EWU, 2024

• Best Paper Award, Image Processing and Image Understanding, ‘Digging into Semantics in the Mask’, 2023

• Solvay Scholarship, Received Tuition and Study Incentives for two years (total funding: 50 million KRW), 2022-2024

• Best Paper Award, Image Processing and Image Understanding, ‘Adaptive Confidence Thresholding for Monocular Depth Estimation’, 2020

• Best Paper Award, Image Processing and Image Understanding, ‘Pseudo Label-Guided Multitask Learning for Scene Understanding’, 2020

• Academic Excellence Award, Ewha W. University, 2016-2020

• Major Leadership Scholarship, Ewha W. University, 2016-2020

• Samsung SCSC Scholarship, Ewha W. University, 2017-2020

• First Prize, Campus Creative Grant Olympiad Startup Contest, ‘Panic disorder cognitive behavioral therapy application development’, 2018

• First Prize, New Industry Convergence University Creative Startup Contest, ‘Healing Pet Loss Syndrome using VR’, 2017

• Second Prize, New Industry Convergence University Creative Startup Contest, ‘VR Exhibition to visit underprivileged children’, 2016

Academic Services

• Conference Reviewer
  : CVPR (2022 - present), ICCV (2023 - present), ECCV (2024 - present)

Mentoring Experiences

• Sungheui Kim (MS student at CVLab, EWU => Currently at KAI)

• Meongeun Kim (MS student at CVLab, EWU => Currently at KAIST)

• Keonhee Ahn (MS student at CVLab, EWU => Currently at KAI)

• Jihye Ahn (MS student at CVLab, EWU)

• Hyejin Park (PhD student at CVLab, EWU)

• Jiyeong Kim (PhD student at CVLab, EWU)

• Seyoung Joung (MS student at CVLab, EWU)

• Minhee Cho (MS student at CVLab, EWU)

• Hayeon Jo (MS student at CVLab, EWU)

• Sumin Park (MS student at CVLab, EWU)

• Sumin Son (MS student at CVLab, EWU => Currently at University of Toronto)

• Sumin Kim (MS student at CVLab, EWU => Currently at University of Toronto)

• Jueun Ko (MS student at CVLab, EWU)

• Hyewon Park (MS student at CVLab, EWU)

• Soyul Lee (MS student at CVLab, EWU)

• Seungmin Beak (MS student at CVLab, EWU)

• Daeun Kim (MS student at CVLab, EWU)

• Yerim So (MS student at CVLab, EWU)