Generative
Medical Imaging
Conditional image synthesis, large-scale visual data, and evaluation of generated outputs.
- My role
- AI research, training
& validation - Research settings
- SNU Hospital iRail Lab
AIVIS - Methods
- GANs · Diffusion · VLMs
Python · PyTorch
Image generation
with evaluation.
My research experience includes conditional chest X-ray synthesis, large-scale medical image data, and generative-model evaluation.
This is a summary of my research practice across roles, rather than a single commercial project. The linked research CV provides the detailed academic and project record.
What I worked on.
- High-resolution synthesisConditional 1024 × 1024 chest X-ray generation using GAN-based methods, alongside comparative generative-model experiments.
- Data & trainingExperience with million-image datasets, preprocessing, synthetic data pipelines, model training, and validation.
- EvaluationCombined expert visual assessment with downstream task performance to assess generated medical images.
- Pathology modelsWorked on pathology detection and classification, foundation-model adaptation, and image-based evaluation.
MICCAI 2025 UNICORN
Pathology Vision Challenge
Recorded in my research and career materials.
A technical basis
for creative work.
Working with synthetic medical images taught me to define what quality means for a task, inspect failure cases, and compare outputs systematically.
I bring that approach to creative production: character consistency, shot intent, and the relationship between generation choices and the finished result.
Python · PyTorch · GANs · Diffusion models · Vision-language models · Git · Linux
Full research CV.
Read the original English CV for the full research, education, and project record.
Read research CV ↗