Seungyong Lee (Ryan)

Seungyong Lee

I am an MS/Ph.D. student at AGI-LAB, KAIST, advised by Prof. Jinwoo Shin. I studied Mathematics and Electrical Engineering at KAIST.

These days I am most interested in physical AI and world models. I dream of a future where robots are present in every home, living alongside people as part of everyday life. Previously, my research focused on computer vision, spanning from perception to generation—understanding the visual world through recognition, and recreating and manipulating it with generative models. I am also passionate about large-scale foundation model training and scalable inference systems from an engineering perspective.

I am also an AI research engineer at NXN Labs, where I lead the development of virtual try-on models and image generation frameworks, taking ownership of the full pipeline—from problem definition and data collection to model design, training, and deployment.

Email / CV / Github / LinkedIn

News

Nov 2026: RoboTryOn is now on arXiv. Check out the project page!
Sep 2026: Joined AGI-LAB, KAIST as an MS/Ph.D. student, advised by Prof. Jinwoo Shin.
Aug 2026: CtrlVTON was accepted to an ECCV 2026 workshop.
Jul 2026: CtrlVTON and the VITON-HD-edit dataset are released.
Oct 2025: Voost was accepted to SIGGRAPH Asia 2025.
Sep 2025: Launched a Hugging Face Space demo for Voost; surpassed 100 K visits in one month and ranked #4 in Trending Spaces.
Jun 2025: Attended CVPR 2025 in Nashville. Great to meet so many of you!
Apr 2025: Our paper was accepted to a CVPR 2025 workshop.
Jun 2024: Attended CVPR 2024 in Seattle.
Jul 2023: Joined NXN Labs as a founding AI research engineer.
Jan 2023: Joined Lunit as an AI Research Scientist Intern.

Publications

	*RoboTryOn: Synthesizing Dexterous Robot Demonstrations from Human Videos via Virtual Try-On Seungyong Lee, Suhyeok Jang, Minyoung Kim, Jinwoo Shin arXiv*, 2026 paper \| project page \| demo \| We treat human-to-robot translation as virtual try-on, turning egocentric human videos into robot demonstrations that teach policies novel tasks without any teleoperation.
	*CtrlVTON: Controllable Virtual Try-On via Visual-Instance-Prompt Segmentation Seungyong Lee, Hyun Jun Jang, Sangoh Kim, Sungjoon Park* ECCV Workshop, 2026 paper \| huggingface \| We recast virtual try-on as image editing with pixel-level mask control, enabling fine-grained control over garment size, style, and placement.
	*Voost: A Unified and Scalable Diffusion Transformer for Bidirectional Virtual Try-On and Try-Off Seungyong Lee, Jeong-gi Kwak* SIGGRAPH Asia, 2025 paper \| project page \| demo \| (#4 trending HF space) We introduced a unified and scalable diffusion transformer for jointly learning virtual try-on and try-off, enabling robust garment–target correspondence.
	Unified Diffusion Transformer for Bidirectional Virtual Try-On and Try-Off Seungyong Lee, Jeong-gi Kwak CVPR 2025 Workshop — AI for Creative Visual Content Generation, Editing and Understanding PDF We proposed a unified diffusion transformer that performs both virtual try-on and try-off, achieving state-of-the-art results on both tasks.
	Fashion Style Editing with Generative Human Prior Chaerin Kong, Seungyong Lee, Soohyeok Im, Wonsuk Yang arXiv:2404.01984 We achieve high-fidelity text-driven fashion style editing in a compute-efficient manner by leveraging a generative human prior.

Work Experiences

	NXN Labs AI Research Engineer, Founding Member (Jul 2023 – Present) Led research and engineering initiatives in virtual try-on and virutal human for large-scale applications. Designed scalable data and training pipelines for multi-node GPU clusters. Developed and maintained a k8s inference architecture serving multiple model endpoints. Built a VLM-driven automation pipeline for image–metadata preprocessing and database ingestion, and defined new research directions in depth-aware segmentation and multi-context image editing. Supervised and mentored four engineer interns, fostering collaboration across model, data, and deployment tasks.
	Lunit AI Research Scientist Intern (Jan 2023 – Jul 2023) Worked on medical image recognition using computer vision techniques, focusing on tumor cell detection and segmentation in pathological images with Vision Transformers.
	Koh Young Technology R&D Center AI Engineer Intern (Aug 2020 – Feb 2021) Participated in an anomaly detection project for inspecting defects on semiconductor substrates and developed data labeling and real-time inference tools.

Template from Jon Barron. Thank you for sharing!