Ph.D. Student in Computer Science & AI · Dongguk University

Junseo Park

I study how diffusion models generate images — and how to make them controllable and interpretable.

I am a Ph.D. student at Dongguk University, advised by Prof. Hyeryung Jang. My research focuses on computer vision, particularly image generation with diffusion models. I am fascinated by the powerful prior of large diffusion models and work on explainable AI (XAI) for generative models and its applications to image editing, style personalization, and optimization.

  • Generative AI
  • Explainable AI
  • Image Editing
  • Multi-Modal Learning
Photo of Junseo Park
D-ION Lab, advised by Prof. Hyeryung Jang Seoul, Korea
ICLR 2025First-author paper (I2AM)
3Papers under review (CVPR ×2, ACM TOMM)
4.50 / 4.50M.S. GPA
115Paper reviews in my notes

News

  1. Received my M.S. and started my Ph.D. at Dongguk University (advisor Prof. Hyeryung Jang).
  2. Our paper StyleForge was published in Applied Sciences (MDPI).
  3. I2AM was accepted to ICLR 2025.
  4. Presented StyleBoost at ICTC 2023.

Publications

Preprint 2026 Preprint

ART-VITON: Measurement-Guided Latent Diffusion for Artifact-Free Virtual Try-On

Junseo Park and Hyeryung Jang

Preprint, 2026

Measurement-guided latent diffusion for artifact-free virtual try-on.

Under review 3 papers currently under review — CVPR ×2, ACM TOMM.

Projects

2024.03 – 2024.06 · Individual project

Robust QA with DistilBERT

Improved out-of-domain question answering (F1 48.41 → 50.31, EM 31.94 → 35.12) with QA-aware text augmentation and AutoML.

  • NLP
  • DistilBERT
  • Data Augmentation

2023.03 – 2023.11 · Individual project

Reinforcement Learning Practice

From-scratch implementations of DQN, DDQN, PPO, A2C/A3C and DDPG, tested on Gym, Unity ML-Agents and a UAV positioning task.

  • Reinforcement Learning
  • PyTorch
  • Unity

2023.03 – 2023.07 · Team project (Capstone Design, 4 members)

Consistent Web Novel Illustration Generation

An AI service that keeps web-novel characters consistent across illustrations by binding each character to a token with DreamBooth.

Role: Built the AI server and trained a DreamBooth-based personalization model for consistent character generation.

  • Stable Diffusion
  • DreamBooth
  • Personalization

Education

  1. Ph.D. in Computer Science and Artificial Intelligence Dongguk University · 2026.03 – Present Advisor — Prof. Hyeryung Jang
  2. M.S. in Computer Science and Artificial Intelligence Dongguk University · 2023.09 – 2026.03 GPA 4.50 / 4.50 · Advisor — Prof. Hyeryung Jang
  3. B.E. in Information and Communication Engineering Dongguk University · 2018.03 – 2023.09 GPA 4.21 / 4.50

Skills

Languages
PythonCJavaSQLHTML & CSS
Technologies
PyTorchFlaskMySQLOpenCVInkscapeGit
Theory — my paper notes 115 paper reviews across 20 research fields, plus theory notes and lecture summaries. Open the notes →

Contact

I am happy to talk about research collaborations and opportunities.

mki730@dgu.ac.kr