Jeonghyeok Do

About

I am a postdoctoral researcher at the Video and Image Computing Lab (VIC Lab), KAIST, where I also received my Ph.D. in Electrical Engineering (Aug. 2025) under the supervision of Prof. Munchurl Kim. I currently lead an NRF/MSIT-funded Sejong Science Fellowship project on AI-based satellite image visualization and analysis.

My research centers on generative models for remote sensing — SAR-to-EO image translation, PAN-sharpening, and satellite image super-resolution — and on efficient skeleton-based action recognition. My current interests extend to remote sensing foundation models, skeleton foundation models, and motion generation.

Research Lines

Remote Sensing

U-Know-DiffPANCVPR 2025PAN-sharpening with uncertainty-aware knowledge distillation
PAN-CrafterICCV 2025PAN-sharpening with modality-consistent alignment
C-DiffSETIEEE TCSVT 2026SAR-to-EO translation with latent diffusion
GeoCore-9BNeurIPS 2026Geo-aware generative foundation model for Earth observation
ReFlowSETarXiv 2026SAR-to-EO translation with latent flow matching

GeoSETarXiv 2026Generalist foundation model for SAR-to-EO translation

GeoCRarXiv 2026Generalist cloud removal prior

Skeleton & Motion

SkateFormerECCV 2024Skeletal-temporal transformer for action recognition
TDSMICCV 2025Diffusion-powered skeleton-text alignment for zero-shot recognition
SLiMNeurIPS 2026Compact-token masked feature prediction for skeleton representations
SOfAarXiv 2026Cross-sensor skeleton foundation model

MotionMaestroarXiv 2026Masked tokenization for unified motion generation

News

2026.09Two papers (GeoCore-9B, SLiM) accepted to NeurIPS 2026.
2026.06AVSR-Diff accepted to ECCV 2026.
2026.06C-DiffSET accepted to IEEE TCSVT; code released on GitHub.
2026.03Started as Principal Investigator of the Sejong Science Fellowship (NRF & MSIT).
2025.09Started as a Postdoctoral Researcher at KAIST.
2025.07Two papers (TDSM, PAN-Crafter) accepted to ICCV 2025.
2025.02U-Know-DiffPAN accepted to CVPR 2025.
2024.07SkateFormer accepted to ECCV 2024.

Publications

International Conferences & Journals

NeurIPS 2026

GeoCore-9B: Towards Geo-Aware Generative Foundation Models in Earth Observation

Jeonghyeok Do, Munchurl Kim

Advances in Neural Information Processing Systems, 2026

arXiv project code

NeurIPS 2026

Less is More: Compact-Token Masked Feature Prediction for Skeleton Representation Learning

Jeonghyeok Do, Yun Chen, Geunhyuk Youk, Munchurl Kim

Advances in Neural Information Processing Systems, 2026

arXiv project code

ECCV 2026

AVSR-Diff: Scale-Agnostic Diffusion Priors for Temporally Consistent Arbitrary-Scale Video Super-Resolution

Geunhyuk Youk, Jeonghyeok Do, Dayeon Kim, Jihyong Oh†, Munchurl Kim†

† Co-corresponding authors · European Conference on Computer Vision, 2026

paper project code video

TCSVT 2026

C-DiffSET: Leveraging Latent Diffusion for SAR-to-EO Image Translation with Confidence-Guided Reliable Object Generation

Jeonghyeok Do, Jaehyup Lee, Seungchul Lee, Munchurl Kim

IEEE Transactions on Circuits and Systems for Video Technology, 2026

paper project code

ICCV 2025

Bridging the Skeleton-Text Modality Gap: Diffusion-Powered Modality Alignment for Zero-shot Skeleton-based Action Recognition

Jeonghyeok Do, Munchurl Kim

International Conference on Computer Vision, 2025

paper supp project code video

ICCV 2025

PAN-Crafter: Learning Modality-Consistent Alignment for PAN-Sharpening

Jeonghyeok Do, Sungpyo Kim, Geunhyuk Youk, Jaehyup Lee†, Munchurl Kim†

† Co-corresponding authors · International Conference on Computer Vision, 2025

paper supp project code video

CVPR 2025

U-Know-DiffPAN: An Uncertainty-aware Knowledge Distillation Diffusion Framework with Details Enhancement for PAN-Sharpening

Sungpyo Kim, Jeonghyeok Do, Jaehyup Lee†, Munchurl Kim†

† Co-corresponding authors · IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2025

paper supp project code

ECCV 2024

SkateFormer: Skeletal-Temporal Transformer for Human Action Recognition

Jeonghyeok Do, Munchurl Kim

European Conference on Computer Vision, 2024

paper supp project code video

VCIP 2023

MorphVAD: Efficient Video Anomaly Detection Using Morphological Transformation

Jeonghyeok Do, Munchurl Kim

IEEE International Conference on Visual Communications and Image Processing, 2023

ICCAS 2022

Multi-modal Transformer for Indoor Human Action Recognition

Jeonghyeok Do, Munchurl Kim

22nd International Conference on Control, Automation and Systems, IEEE, 2022

IEEE Access 2021

Pseudo-Supervised Learning for Semantic Multi-Style Transfer

Saehun Kim, Jeonghyeok Do, Munchurl Kim

IEEE Access, vol. 9, pp. 7930–7942, 2021

IEEE Access 2020

Learning-based JND-directed HDR Video Preprocessing for Perceptually Lossless Compression with HEVC

Sehwan Ki, Jeonghyeok Do, Munchurl Kim

IEEE Access, vol. 8, pp. 228605–228618, 2020

CVPRW 2020

NTIRE 2020 Challenge on NonHomogeneous Dehazing

Codruta O. Ancuti, Cosmin Ancuti, Florin-Alexandru Vasluianu, Radu Timofte, et al. (incl. Jeonghyeok Do)

IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2020

Preprints

arXiv 2026

MotionMaestro: Masked Tokenization for Unified Motion Generation

Yun Chen, Munchurl Kim†, Jeonghyeok Do†

† Co-corresponding authors · arXiv preprint, 2026

arXiv project code

arXiv 2026

GeoSET: Generalist Foundation Model for SAR-to-EO Image Translation

Jeonghyeok Do, Munchurl Kim

arXiv preprint, 2026

arXiv project code

arXiv 2026

GeoCR: Learning a Generalist Cloud Removal Prior from Heterogeneous Observations

Jeonghyeok Do, Munchurl Kim

arXiv preprint, 2026

arXiv project code

arXiv 2026

ReFlowSET: Representation-Aligned Latent Flow Matching for SAR-to-EO Image Translation

Jeonghyeok Do, Seungchul Lee, Munchurl Kim

arXiv preprint, 2026

arXiv project code

arXiv 2026

One for All: Generalist Foundation Model for Cross-Sensor Skeleton Representation Learning

Jeonghyeok Do, Yun Chen, Munchurl Kim

arXiv preprint, 2026

arXiv project code

Under Review

Under Review 2026

SAREO-FM: Decoupled Semantic Supervision for SAR–EO Foundation Models

Jeonghyeok Do, Munchurl Kim

Under review, 2026

Under Review 2026

What Makes a Good Motion Tokenizer for 3D Human Motion Generation?

Yun Chen*, Jeonghyeok Do*, Munchurl Kim

* Equal contribution · Under review, 2026

Experience

2026.03 – present Principal Investigator, Sejong Science Fellowship AI-based Satellite Image Visualization and Analysis · funded by NRF & MSIT
2025.09 – present Postdoctoral Researcher, KAIST Information & Electronics Research Institute
2019.03 – 2025.08 Research Assistant, KAIST Government- and industry-funded projects: ADD, LG Electronics, SK Hynix, Stellarvision

Education

2021.03 – 2025.08 Ph.D., School of Electrical Engineering, KAIST Thesis: A Study on Efficient Skeletal-Temporal Transformer for Skeleton-based Action Recognition · Advisor: Prof. Munchurl Kim
2019.03 – 2021.02 M.S., School of Electrical Engineering, KAIST Thesis: Effective Labeling on Unlabeled Data for Imbalanced Training Dataset · Advisor: Prof. Munchurl Kim
2015.03 – 2019.02 B.S., School of Electrical Engineering, KAIST Double major in Mathematical Sciences · Cum Laude

Academic Service

Conference reviewer — AAAI (2027); NeurIPS, AAAI, BMVC (2026); ICCV, CVPR (2025); VCIP (2023–2025)
Journal reviewer — IJCV, IEEE TIP, TNNLS, TMM, ACM TOMM

Invited Talks

Deep Learning-based Multi-modal Satellite Image Fusion
Prof. Jungho Im’s Lab, UNIST · Feb. 2026

Honors & Awards

Government Scholarship for Ph.D. students, 2021–2025
KAIST Scholarship for M.S. students, 2019–2021
National Scholarship for undergraduates, 2015–2019
KAIST President Award, 2015

Technology Transfer

Generative AI-based SAR-to-EO Image Translation
transferred to Stellarvision · Apr. 2025