Seokeon Choi

I am a Staff Research Engineer in the Personalization of generative AI and On-device learning team at Qualcomm AI Research. Before joining Qualcomm, I completed my Ph.D. in Electrical Engineering at KAIST, advised by Prof. Changick Kim. During my Ph.D., I was a Research Intern in the TensorFlow Model Optimization team at Google, and a Visiting Researcher at CMU under the guidance of Prof. Alex Hauptmann.

Email: bismex (at) gmail.com / seokchoi (at) qti.qualcomm.com

Research Experiences

Qualcomm AI Research, Seoul, Korea
- Staff Research Engineer (Nov. 2023 - present)
- Senior Research Engineer (Sep. 2021 - Nov. 2023)
Google, Remote, Korea
- Research Intern (Jun. 2021 - Sep. 2021)
Carnegie Mellon University, Pittsburgh, PA
- Visiting Researcher (Jan. 2020 - Jul. 2020)
KAIST, Daejeon, Korea
- Research Assistant (Mar. 2015 - Jun. 2021)

Education

PhD in EE, KAIST, 2021 (Thesis: Person Re-Identification to Bridge Domain Gaps)
MS in EE, KAIST, 2017 (Thesis: Robust Model-based Gait Recognition via Candidate Selection and Pose-aware Decision Fusion)
BS in EEE, SKKU, 2015

News

[new] One paper accepted to ICCVW'25 (Oral)
[new] Two papers accepted to ICCV'25
One paper accepted to NeurIPS'24
One paper accepted to ECCV'24
I've been promoted to Staff Research Engineer at Qualcomm
One paper accepted to CVPR'23
One paper accepted to CVPRW'23
One paper accepted to TPAMI'22
One paper accepted to ECCV'22
Two papers accepted to CVPR'21
One paper accepted to CVPRW'21
One paper accepted to ICCV'21
One paper accepted to CVPR'20
Two papers accepted to ECCVW'20
One paper accepted to CVPR'19
Two papers accepted to ECCVW'18

Research Interest

My research interests lie in computer vision and machine learning. The purpose of my research is to bridge the gap between cutting-edge machine learning studies and real-world problems, thereby providing practical solutions. With this goal in mind, my current research focuses on (1) Generative AI and Multimodality, (2) Transferability and Generalizability, and (3) Computer Vision Applications (Human Understanding and Machine Perception).

In particular, I am dedicated to researching text-to-image/text-to-video diffusion models with a focus on personalization and efficient training/inference.

Career Vision

With the rapid advancement of generative AI, the potential for groundbreaking innovation is limitless. My goal is to pioneer novel methodologies that will redefine the AI landscape, creating significant business impact in the process. Drawing on my diverse experience across multiple research domains, I am uniquely positioned as a generalist capable of applying cutting-edge research to a variety of challenges, ensuring a broader and more impactful contribution to the AI ecosystem.

Research Timeline

Selected Publications

Research Categories

Generative AI and Multimodality (2019 ~ present):
- Synthetic data generation [CVPR'23, CVPR'20, CVPR'19, CVPRW'23]
- Text-to-image/video editing and generation [NeurIPS'24, ICCV'25a, Preprint-a, Preprint-b, Preprint-c, Preprint-d]
- Parameter-Efficient Fine-Tuning (PEFT) [ICCV'25b]
Transferability and Generalizability (2019 ~ 2024):
- Domain generalization [ECCV'24, CVPR'23, CVPR'21a]
- Domain adaptation [ECCV'22, CVPR'19]
- Cross-modal learning [CVPR'20]
- Transfer learning [CVPRW'23]
- Few-shot learning [CVPR'21b]
Human Understanding (2015-2022):
- Person re-identification [CVPR'21a, CVPR'20, arXiv'19]
- Pose estimation [CVPRW'21]
- Action detection [TPAMI'22]
- Gait recognition (Biometrics) [TIFS'19]
Machine Perception (2015~2021):
- Object detection [CVPR'19]
- Object tracking [BMVC'19, ECCVW'20, ECCVW'18]
- Object segmentation [WACV'20]
- 3D reconstruction [ICCV'21]
- Challenge report related to perception tasks [ECCVW'20, ECCVW'18]
Ongoing Research (Present~):

- Multi-subject / Multi-part Human Generation
- Memory-efficient Personalization of Diffusion Models
- Image Generation with Reinforcement Learning & Reasoning
- Multimodal LLMs with Diffusion Models
- Personalization of Consistency Models (for Faster Inference)

- Visual Reinforcement Fine-Tuning
- Personalization of MLLMs
- Zeroth-order Optimization for Fine-tuning (LLM/LVM/LMM)
- Low-precision Fine-tuning (LLM/LVM/LMM)

Research Highlights

Efficient Fine-Tuning Diffusion Transformer via Low-Resolution Sampling and Layer Skipping

Sunghyun Park, Jeongho Kim, Hyoungwoo Park, Debasmit Das, Sungrack Yun, Munawar Hayat, Jaegul Choo, Fatih Porikli, Seokeon Choi

Under review

[Preprint coming soon 🥳]

MultiHuman-Testbench: Benchmarking Image Generation for Multiple Humans

Shubhankar Borse, Seokeon Choi, Sunghyun Park, Jeongho Kim, Shreya Kadambi, Risheek Garrepalli, Sungrack Yun, Munawar Hayat, Fatih Porikli

Under review

[Preprint] [code&dataset]

From Wardrobe to Canvas: Wardrobe Polyptych LoRA for Part-level Controllable Human Image Generation

Jeongho Kim, Sunghyun Park, Hyoungwoo Park, Sungrack Yun, Jaegul Choo, Seokeon Choi

Under review

[Preprint]

Memory-Efficient Personalization of Text-to-Image Diffusion Models via Selective Optimization Strategies

*Seokeon Choi, *Sunghyun Park, Hyoungwoo Park, Jeongho Kim, Sungrack Yun (*Equal contribution)

IEEE/CVF International Conference on Computer Vision (ICCV)
Workshop on Representation Learning with Very Limited Resources (LIMIT), 2025

Oral presentation

[Paper]

Steering Guidance for Personalized Text-to-Image Diffusion Models

*Sunghyun Park, *Seokeon Choi, Hyoungwoo Park, Sungrack Yun (*Equal contribution)

IEEE/CVF International Conference on Computer Vision (ICCV), 2025

[Paper]

ConsNoTrainLoRA: Data-driven Weight Initialization of Low-rank Adapters using Constraints

*Debasmit Das, *Hyoungwoo Park, Munawar Hayat, Seokeon Choi, Sungrack Yun, Fatih Porikli (*Equal contribution)

IEEE/CVF International Conference on Computer Vision (ICCV), 2025

[Paper]

Hollowed Net for On-Device Personalization of Text-to-Image Diffusion Models

Wonguk Cho, Seokeon Choi, Debasmit Das, Matthias Reisser, Taesup Kim, Sungrack Yun, Fatih Porikli

Neural Information Processing Systems (NeurIPS), 2024

[paper]

Feature Diversification and Adaptation for Federated Domain Generalization

Seunghan Yang, Seokeon Choi, Hyunsin Park, Sungha Choi, Simyung Chang, Sungrack Yun

European Conference on Computer Vision (ECCV), 2024

[paper]

Progressive Random Convolutions for Single Domain Generalization

Seokeon Choi, Debasmit Das, Sungha Choi, Seunghan Yang, Hyunsin Park, Sungrack Yun

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023

[paper] [supp] [video] [slide] [poster]

Neural Transformation Network to Generate Diverse Views for Contrastive Learning

Taekyung Kim, Debasmit Das, Seokeon Choi, Minki Jeong, Seunghan Yang, Sungrack Yun, Changick Kim

IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshop (CVPRW), 2023

[paper] [supp] [poster]

Improving Test-Time Adaptation via Shift-agnostic Weight Regularization and Nearest Source Prototypes

Sungha Choi, Seunghan Yang, Seokeon Choi, Sungrack Yun

European Conference on Computer Vision (ECCV), 2022

[paper]

Learning to Discriminate Information for Online Action Detection: Analysis and Application

Sumin Lee, Hyunjun Eun, Jinyoung Moon, Seokeon Choi, Yoonhyung Kim, Chanho Jung, Changick Kim

IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 2022

[paper] [code]

Just a Few Points are All You Need for Multi-view Stereo: A Novel Semi-supervised Learning Method for Multi-view Stereo

Taekyung Kim, Jaehoon Choi, Seokeon Choi, Dongki Jung, Changick Kim

IEEE/CVF International Conference on Computer Vision (ICCV), 2021

[paper] [supp]

Meta Batch-Instance Normalization for Generalizable Person Re-Identification

Seokeon Choi, Taekyung Kim, Minki Jeong, Hyoungseob Park, Changick Kim

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021

[paper] [supp] [code]

Few-shot Open-set Recognition by Transformation Consistency

Minki Jeong, Seokeon Choi, Changick Kim

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021

[paper] [supp] [code]

MobileHumanPose: Toward Real-time 3D Human Pose Estimation in Mobile Devices

Sangbum Choi, Seokeon Choi, Changick Kim

IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshop (CVPRW), 2021

[paper] [code]

Robust Long-Term Object Tracking via Improved Discriminative Model Prediction

Seokeon Choi, Junhyun Lee, Yunsung Lee, Alex Hauptmann

European Conference on Computer Vision Workshop (ECCVW), 2020

[paper] [code] [video(1min, ENG)] [video(12min, ENG)]

Hi-CMD: Hierarchical Cross-Modality Disentanglement for Visible-Infrared Person Re-Identification

Seokeon Choi, Sumin Lee, Youngeun Kim, Taekyung Kim, Changick Kim

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020

[paper] [supp] [code] [video(1min, ENG)] [video(5min, KOR)] [poster] [slide(short)] [slide(long)]

RPM-Net: Robust Pixel-Level Matching Networks for Self-Supervised Video Object Segmentation

Youngeun Kim, Seokeon Choi, Hankyeol Lee, Taekyung Kim, Changick Kim

IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2020

[paper] [supp] [video(oral presentation)]

Bilinear Siamese Networks with Background Suppression for Visual Object Tracking

Hankyeol Lee, Seokeon Choi, Youngeun Kim, Changick Kim

British Machine Vision Conference (BMVC), 2019 (Spotlight)

[paper]

Diversify and Match: A Domain Adaptive Representation Learning Paradigm for Object Detection

Taekyung Kim, Minki Jeong, Seunghyeon Kim, Seokeon Choi, Changick Kim

IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019

[paper] [supp] [code]

Skeleton-based Gait Recognition via Robust Frame-level Matching

Seokeon Choi, Jonghee Kim, Wonjun Kim, Changick Kim

IEEE Transactions on Information Forensics and Security (TIFS), 2019

[paper] [code] [dataset]

A Memory Model based on the Siamese Network for Long-term Tracking

*Hankyeol Lee, *Seokeon Choi, Changick Kim (*Equal contribution)

European Conference on Computer Vision Workshop (ECCVW), 2018

[paper] [code]

US patents

Seokeon Choi, Juntae Lee, Jaewon Choi, “Privacy-Aware Multi-Modal Generative Autoreply,” U.S. Patent Application No. 18/454,456. [link]
Seunghan Yang, Seokeon Choi, Hyunsin Park, Sungha Choi, Sungrack Yun, “Client-agnostic learning and zero-shot adaptation for federated domain generalization,” U.S. Patent Application No. 18/238,998. [link]
Seokeon Choi, Sungha Choi, Seunghan Yang, Hyunsin Park, Debasmit Das, and Sungrack Yun, “Semantic-aware random style aggregation for single domain generalization,” U.S. Patent Application No. 18/157,723. [link]
Sungha Choi, Seunghan Yang, Seokeon Choi, and Sungrack Yun, “Test-time adaptation with unlabeled online data,” U.S. Patent Application No. 18/086,586. [link]

Awards, Honors, and Scholarship

CVPR2021 Doctoral Consortium Presenter (Mentor: Tong Xiao from Meta) - 2021
Academic scholarship funded by KAIST Chul-Hi Han Augustino scholarship foundation (880$) - 2021
27th Samsung Humantech Paper Award, Silver Prize, Top 2% (6,274$) - 2021
26th Samsung Humantech Paper Award, Finalist, Top 8% - 2020
Completion of Deep Learning class (deeplearning.ai) at Coursera - 2020
Google Developer Machine Learning Bootcamp - 2020 | 2021
Carnegie Mellon University, Intensive AI Program fully supported by Korean government (43,435$) - 2020
CVPR 2020 AI City challenge 16th prize winner (vehicle re-identification part) - 2020
ECCV 2020 VOT2020-LT challenge 5th prize winner - 2020
ECCV 2018 VOT2018-LT challenge 3rd prize winner - 2018
SAIT (Samsung Advanced Institute of Technology) Industry-Academic Scholarship Students (49,167$) - 2021
Summa Cum Laude, Sung-Kyun-Kwan University (GPA: 4.41/4.50) (99.0%) - 2015
Best Work/Thesis Award, Sung-Kyun-Kwan University - 2014
Kepco E&C, Power Engineering School Winter Camp - 2013
Dean's List, Sung-Kyun-Kwan University - 2012 | 2013
National Science & Technology Scholarship (16,329$) - 2012 | 2013 | 2014
Academic scholarship, Sung-Kyun-Kwan University (10,633$) - 2009 | 2012
Served and discharged from ROK Army as Sergeant - 2010 | 2011
Mensa (The High IQ society) - 2012 - Present

Professional Activities

Reviewer of International Conferences

Annual Conference on Neural Information Processing Systems (NeurIPS) - 2025
IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) - 2023 | 2024 | 2025
IEEE/CVF International Conference on Computer Vision (ICCV) - 2023 | 2025
European Conference on Computer Vision (ECCV) - 2024
ACM Multimedia (ACMMM) - 2025
IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) - 2025

Reviewer of International Journals

International Journal of Computer Vision - 2023 | 2024 | 2025
IEEE Transactions on Neural Networks and Learning Systems - 2022 | 2023
IEEE Transactions on Image Processing - 2023
IEEE Transactions on Multimedia - 2020 | 2021 | 2022 | 2023
IEEE Transactions on Information Forensics and Security - 2022
IEEE Transactions on Circuits and Systems for Video Technology - 2022
IEEE Transactions on Geoscience and Remote Sensing - 2021
IEEE Internet of Things Journal - 2021
IEEE Signal Processing Letters - 2021 | 2022
IET Computer Vision - 2022
Neurocomputing, Elsevier - 2020

Teaching Assistant

EE209, Programming Structure for Electrical Engineering (30 months) - 2015 | 2016 | 2019
EE838, Image Engineering: Advanced Feature extraction and 3D reconstruction (6 months) - 2018
EE734, Image Understanding (6 months) - 2018
EE838, Image Engineering: Feature matching and 3D reconstruction (6 months) - 2017
EE838, Image Engineering: Dimension Reduction and Optimization Techniques (6 months) - 2016
EE405A, Electronics Design Lab. <RoboCam> (6 months) - 2016
EE202, Signals and Systems (6 months) - 2015

Invited Talks

"Solving the Domain Shift Problem in Deep Learning" CVPR 2020 | CVPR 2021 | ECCV 2022 :: Korea Advanced Institute of Science and Technology (KAIST), hosted by Center for Security Technology Research - 2023
“Person Re-Identification to Bridge Domain Gaps” CVPR 2020 | CVPR 2021 :: Electronics and Telecommunications Research Institute (ETRI), hosted by Hyung-Il Kim - 2020
“Introduction to person re-identification” CVPR 2020 | CVPR 2021 :: Samsung Advanced Institute of Technology (SAIT), hosted by Jae-Joon Han - 2020
“Introduction to long-term object tracking” ECCVW 2018 | ECCVW 2020 :: Samsung Advanced Institute of Technology (SAIT), hosted by Jae-Joon Han - 2020

Page updated

Google Sites

Report abuse