Patrick Kwon (권용재)
I am a first-year research assistant at the Center for Research in Computer Vision, UCF, working with Dr. Chen Chen. Previously, I worked as a full-time research scientist at Naver Webtoon Corp. studying computer vision, graphics, and AI. I have also worked with Professor Hanbyul Joo and SNU VCLab in solving 3D reconstruction and affordance problems. I received my Master’s degree at Columbia University studying machine learning and human-computer interaction within augmented reality.
My main research goal is in embedding human prior knowledge within generative AI systems. Specifically, I aspire to create AI systems that has a human level understanding towards visual context, which in turn will pave way for novel solutions in the field of computer vision. For instance, how can we aid video generation models to better understand human-object interaction? And how can we utilize such models in helping content creators and improving robotics systems? These are just few of the questions I hope to discover answers for.
news
| Sep 26, 2026 | Our paper FactorizedHMR: A Hybrid Framework for Video Human Mesh Recovery has been accepted as a poster at NeurIPS 2026! |
|---|
selected publications
-
FactorizedHMR: A Hybrid Framework for Video Human Mesh RecoveryIn Advances in Neural Information Processing Systems, Aug 2026A two-stage human mesh recovery (HMR) framework that uses a factorized representation and processing of human body poses. Accepted as a poster at NeurIPS 2026. -
Fall Risk and Gait Analysis in Community-Dwelling Older Adults using World-Spaced 3D Human Mesh RecoveryIn Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW): Computer Vision for Biomechanics Workshop (CVBW), Jun 2026Case Study for utilizing 3D Human Mesh Recovery (HMR) research for analyzing gait for the elderly population -
DreamingComics: A Story Visualization Pipeline via Subject and Layout Customized Generation using Video ModelsIn Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Dec 2025A story visualization pipeline that repurposes video generative models and LLMs to produce identity-consistent, multi-panel comics with spatial layout control. -
GraspDiffusion: Synthesizing Realistic Whole-body Hand-Object InteractionIn Proceedings of the Winter Conference on Applications of Computer Vision (WACV), Oct 2025Generates realistic full-bodied HOI poses and images controlled by a given 3D object via a two-stage synthesis pipeline. -
Beyond the Contact: Discovering Comprehensive Affordance for 3D Objects from Pre-trained 2D Diffusion ModelsIn Computer Vision – ECCV 2024, Sep 2024Discovers rich 3D affordance relationships based on image generative models and their prior knowledge. Oral presentation at ECCV 2024. -
Chupa: Carving 3D Clothed Humans from Skinned Shape Priors using 2D Diffusion Probabilistic ModelsIn Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), Jul 2023Realistic 3D clothed human reconstruction method leveraging image diffusion models and skinned shape priors. Oral presentation at ICCV 2023.