Seungwook Kim

Research Scientist, PhD

Or you can also call me Wookie. I am a Tech Lead in the research team at GenGenAI, where I work on generating data with AI, for AI. I received my integrated MSc/PhD from POSTECH Computer Vision Lab, advised by Prof. Minsu Cho. My research spans any-modality-to-any-modality generation, world models, and self-improving AI systems (ongoing), {text,image}-to-{image,3D,video} generation (CorrespondentDream, RapidMV, FreeAction, SOLACE), visual correspondence (TransforMatcher, HCCNet, MambaMatcher, DHVR), and 2D/3D equivariance (SeLCA, CHOIR, RIST).

swookie.kim (at) gmail (dot) com wookiekim (at) postech (dot) ac (dot) kr

News

2026-06 Recognized as Outstanding Reviewer for CVPR 2026.
2026-04 Started as Tech Lead / Research Scientist at GenGenAI, Seoul, working on generating data with AI, for AI.
2026-03 Gave an invited talk on "Improving the Fidelity of Diffusion Generative Models with Intrinsic Self Guidance" at Chung-Ang University, hosted by Prof. Jongmin Lee.
2026-02 1 paper on text-to-image generation (SOLACE) accepted to CVPR 2026.
2025-06 1 paper on semantic correspondence (MambaMatcher) accepted to ICCV 2025 (Findings, Oral).
2025-06 1 paper on text-to-3D generation (Training-Free SDS) accepted to CVPR 2025 Workshop on AI4CC.
2025-03 Recognized as Outstanding Reviewer for ICCV 2025.
2025-01 1 paper on multi-view generation (RapidMV) accepted to WACV 2026.
2024-06 Started PhD internship at ByteDance Seed Team, CA, USA.
2024-02 2 papers on text-to-3D generation (CorrespondentDream) and 3D equivariance (RIST) accepted to CVPR 2024.
2024-01 Received POSTECHIAN Fellowship (Innovation) 2024.
2023-09 Started PhD internship at ByteDance, CA, USA.

Selected Publications

Improving Text-to-Image Generation with Intrinsic Self-Confidence Rewards

Seungwook Kim, Minsu Cho

The IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR) 2026

Paper Code Project Blog

@inproceedings{kim2026intrinsic,
  title={Improving Text-to-Image Generation with Intrinsic Self-Confidence Rewards},
  author={Kim, Seungwook and Cho, Minsu},
  booktitle={CVPR},
  year={2026}
}

FreeAction: Training-Free Techniques for Enhanced Fidelity of Trajectory-to-Video Generation

Seungwook Kim, Seunghyeon Lee, Minsu Cho

2025 Conference on Robot Learning @ Seoul, South Korea - Workshop on Learning to Simulate Robot Worlds

Paper Code Project Blog

@inproceedings{kim2025freeaction,
  title={FreeAction: Training-Free Techniques for Enhanced Fidelity of Trajectory-to-Video Generation},
  author={Kim, Seungwook and Lee, Seunghyeon and Cho, Minsu},
  booktitle={CoRL Workshop},
  year={2025}
}

RapidMV: Leveraging Spatio-Angular Representations for Efficient and Consistent Text-to-Multi-View Synthesis

Seungwook Kim, Yichun Shi, Kejie Li, Minsu Cho, Peng Wang

IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) 2026 @ Arizona

Paper Code Project Blog

@inproceedings{kim2026rapidmv,
  title={RapidMV: Leveraging Spatio-Angular Representations for Efficient and Consistent Text-to-Multi-View Synthesis},
  author={Kim, Seungwook and Shi, Yichun and Li, Kejie and Cho, Minsu and Wang, Peng},
  booktitle={WACV},
  year={2026}
}

Similarity-Aware Selective State-Space Modeling for Semantic Correspondence

Seungwook Kim, Minsu Cho

The IEEE / CVF International Conference on Computer Vision (ICCV) 2025 @ Honolulu, Hawaii - Findings (Oral)

Paper Code Project Blog

@inproceedings{kim2025similarity,
  title={Similarity-Aware Selective State-Space Modeling for Semantic Correspondence},
  author={Kim, Seungwook and Cho, Minsu},
  booktitle={ICCV},
  year={2025}
}

Enhancing 3D Fidelity of Text-to-3D using Cross-View Correspondences

Seungwook Kim, Kejie Li, Xueqing Deng, Yichun Shi, Minsu Cho, Peng Wang

The IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR) 2024 @ Seattle

Paper Code Project Blog

@inproceedings{kim2024correspondentdream,
  title={Enhancing 3D Fidelity of Text-to-3D using Cross-View Correspondences},
  author={Kim, Seungwook and Li, Kejie and Deng, Xueqing and Shi, Yichun and Cho, Minsu and Wang, Peng},
  booktitle={CVPR},
  year={2024}
}

Learning SO(3)-Invariant Semantic Correspondence via Local Shape Transform

Chunghyun Park*, Seungwook Kim*, Jaesik Park, Minsu Cho

The IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR) 2024 @ Seattle

Paper Code Project Blog

@inproceedings{park2024so3,
  title={Learning SO(3)-Invariant Semantic Correspondence via Local Shape Transform},
  author={Park, Chunghyun and Kim, Seungwook and Park, Jaesik and Cho, Minsu},
  booktitle={CVPR},
  year={2024}
}

Efficient Semantic Matching with Hypercolumn Correlation

Seungwook Kim, Juhong Min, Minsu Cho

IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) 2024 @ Big Island

Paper Code Project Blog

@inproceedings{kim2024efficient,
  title={Efficient Semantic Matching with Hypercolumn Correlation},
  author={Kim, Seungwook and Min, Juhong and Cho, Minsu},
  booktitle={WACV},
  year={2024}
}

Stable and Consistent Prediction of 3D Characteristic Orientation via Invariant Residual Learning

Seungwook Kim*, Chunghyun Park, Yoonwoo Jeong, Jaesik Park, Minsu Cho

International Conference on Machine Learning (ICML) 2023 @ Honolulu

Paper Code Project Blog

@inproceedings{kim2023stable,
  title={Stable and Consistent Prediction of 3D Characteristic Orientation via Invariant Residual Learning},
  author={Kim, Seungwook and Park, Chunghyun and Jeong, Yoonwoo and Park, Jaesik and Cho, Minsu},
  booktitle={ICML},
  year={2023}
}

TransforMatcher: Match-to-Match Attention for Semantic Correspondence

Seungwook Kim, Juhong Min, Minsu Cho

The IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR) 2022 @ New Orleans

Paper Code Project Blog

@inproceedings{kim2022transformatcher,
  title={TransforMatcher: Match-to-Match Attention for Semantic Correspondence},
  author={Kim, Seungwook and Min, Juhong and Cho, Minsu},
  booktitle={CVPR},
  year={2022}
}

Honors & Awards

2026 Outstanding Reviewer, CVPR 2026
2025 Outstanding Reviewer, ICCV 2025
2024 Best Paper Award, IPIU 2024 (1 / 311 papers)
2024 Best Paper Candidate, WACV 2024 (Top 0.6%)
2024 POSTECHIAN Fellowship (Innovation) — $6,000
2023-25 Hyundai Chung Mong-Koo Foundation Scholarship — ~$13,000/yr
2022 POSTECHIAN Fellowship (Leadership) — $1,000
2022,23 BK21 Outstanding Research Paper Award — $500
2020 NAVER AI RUSH — 1st runner-up (×2) — $11,000
2015-20 Jigok Scholarship (Full scholarship), POSTECH — ~$24,000

Experiences

GenGenAI, Seoul, South Korea

Generating data with AI, for AI. Building scalable pipelines for synthetic image/video data across modalities (EO/IR) in data-scarce scenarios. Developing self-evaluation methods that use intrinsic signals from generative models, and a foundation model for controllable synthetic data generation.

ByteDance, CA, USA

Worked on improving the efficiency and consistency of {image,text}-to-multiview generation models. Published 1 paper: RapidMV (WACV 2026).

ByteDance, CA, USA

Worked on improving {image,text}-to-3D generation models. Published 2 papers: CorrespondentDream (CVPR 2024) and MultiImageDream (Arxiv 2024).

ACCV, Hanoi, Vietnam

Managed the Microsoft CMT for coordinating paper submission and reviewing process.

Polaris3D, Pohang, South Korea

Implemented the process of retrieving data from Intel Realsense cameras to Jetson Nano in real-time. Merged the two streams of data from two different angles to output a 3D map in real-time.

SK Hynix, Icheon, South Korea

Analyzed the image post-processing algorithms applied to raw images obtained from sensors.

Netmarble, Seoul, South Korea

Developed prior speech-to-3D lip synthesis pipeline to be light-weight (mobile-runnable) using TensorFlow.

Dable, Seoul, South Korea

Analyzed heavy-traffic raw data collected at AWS RedShift using PostgreSQL. Developed batch codes and web crawling system.

Education

POSTECH

Advisor: Prof. Minsu Cho

POSTECH

GPA: 3.70 / 4.3

All Publications

Improving Text-to-Image Generation with Intrinsic Self-Confidence Rewards

Seungwook Kim, Minsu Cho

The IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR) 2026

Paper Code Project Blog

@inproceedings{kim2026intrinsic,
  title={Improving Text-to-Image Generation with Intrinsic Self-Confidence Rewards},
  author={Kim, Seungwook and Cho, Minsu},
  booktitle={CVPR},
  year={2026}
}

FreeAction: Training-Free Techniques for Enhanced Fidelity of Trajectory-to-Video Generation

Seungwook Kim, Seunghyeon Lee, Minsu Cho

2025 Conference on Robot Learning @ Seoul, South Korea - Workshop on Learning to Simulate Robot Worlds

Paper Code Project Blog

@inproceedings{kim2025freeaction,
  title={FreeAction: Training-Free Techniques for Enhanced Fidelity of Trajectory-to-Video Generation},
  author={Kim, Seungwook and Lee, Seunghyeon and Cho, Minsu},
  booktitle={CoRL Workshop},
  year={2025}
}

RapidMV: Leveraging Spatio-Angular Representations for Efficient and Consistent Text-to-Multi-View Synthesis

Seungwook Kim, Yichun Shi, Kejie Li, Minsu Cho, Peng Wang

IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) 2026 @ Arizona

Paper Code Project Blog

@inproceedings{kim2026rapidmv,
  title={RapidMV: Leveraging Spatio-Angular Representations for Efficient and Consistent Text-to-Multi-View Synthesis},
  author={Kim, Seungwook and Shi, Yichun and Li, Kejie and Cho, Minsu and Wang, Peng},
  booktitle={WACV},
  year={2026}
}

Similarity-Aware Selective State-Space Modeling for Semantic Correspondence

Seungwook Kim, Minsu Cho

The IEEE / CVF International Conference on Computer Vision (ICCV) 2025 @ Honolulu, Hawaii - Findings (Oral)

Paper Code Project Blog

@inproceedings{kim2025similarity,
  title={Similarity-Aware Selective State-Space Modeling for Semantic Correspondence},
  author={Kim, Seungwook and Cho, Minsu},
  booktitle={ICCV},
  year={2025}
}

Harnessing the Power of Training-Free Techniques for Text-to-3D Generation via Score Distillation Sampling

Junhong Lee, Seungwook Kim, Minsu Cho

The IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR) 2025 @ Nashville - Workshop on AI4CC

Paper Code Project Blog

@inproceedings{lee2025training,
  title={Harnessing the Power of Training-Free Techniques for Text-to-3D Generation via Score Distillation Sampling},
  author={Lee, Junhong and Kim, Seungwook and Cho, Minsu},
  booktitle={CVPR Workshop},
  year={2025}
}

3D Geometric Shape Assembly via Efficient Point Cloud Matching

Nahyuk Lee*, Juhong Min*, Junha Lee, Seungwook Kim, Kanghee Lee, Jaesik Park, Minsu Cho

The 41st International Conference on Machine Learning (ICML) 2024 @ Vienna

Paper Code Project Blog

@inproceedings{lee2024shape,
  title={3D Geometric Shape Assembly via Efficient Point Cloud Matching},
  author={Lee, Nahyuk and Min, Juhong and Lee, Junha and Kim, Seungwook and Lee, Kanghee and Park, Jaesik and Cho, Minsu},
  booktitle={ICML},
  year={2024}
}

Multi-view Image Prompted Multi-view Diffusion for Improved 3D Generation

Seungwook Kim, Yichun Shi, Kejie Li, Minsu Cho, Peng Wang

Arxiv 2024

Paper Code Project Blog

@misc{kim2024multiimagedream,
  title={Multi-view Image Prompted Multi-view Diffusion for Improved 3D Generation},
  author={Kim, Seungwook and Shi, Yichun and Li, Kejie and Cho, Minsu and Wang, Peng},
  year={2024},
  eprint={2404.17419},
  archivePrefix={arXiv}
}

Enhancing 3D Fidelity of Text-to-3D using Cross-View Correspondences

Seungwook Kim, Kejie Li, Xueqing Deng, Yichun Shi, Minsu Cho, Peng Wang

The IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR) 2024 @ Seattle

Paper Code Project Blog

@inproceedings{kim2024correspondentdream,
  title={Enhancing 3D Fidelity of Text-to-3D using Cross-View Correspondences},
  author={Kim, Seungwook and Li, Kejie and Deng, Xueqing and Shi, Yichun and Cho, Minsu and Wang, Peng},
  booktitle={CVPR},
  year={2024}
}

Learning SO(3)-Invariant Semantic Correspondence via Local Shape Transform

Chunghyun Park*, Seungwook Kim*, Jaesik Park, Minsu Cho

The IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR) 2024 @ Seattle

Paper Code Project Blog

@inproceedings{park2024so3,
  title={Learning SO(3)-Invariant Semantic Correspondence via Local Shape Transform},
  author={Park, Chunghyun and Kim, Seungwook and Park, Jaesik and Cho, Minsu},
  booktitle={CVPR},
  year={2024}
}

Efficient Semantic Matching with Hypercolumn Correlation

Seungwook Kim, Juhong Min, Minsu Cho

IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) 2024 @ Big Island

Paper Code Project Blog

@inproceedings{kim2024efficient,
  title={Efficient Semantic Matching with Hypercolumn Correlation},
  author={Kim, Seungwook and Min, Juhong and Cho, Minsu},
  booktitle={WACV},
  year={2024}
}

Stable and Consistent Prediction of 3D Characteristic Orientation via Invariant Residual Learning

Seungwook Kim*, Chunghyun Park, Yoonwoo Jeong, Jaesik Park, Minsu Cho

International Conference on Machine Learning (ICML) 2023 @ Honolulu

Paper Code Project Blog

@inproceedings{kim2023stable,
  title={Stable and Consistent Prediction of 3D Characteristic Orientation via Invariant Residual Learning},
  author={Kim, Seungwook and Park, Chunghyun and Jeong, Yoonwoo and Park, Jaesik and Cho, Minsu},
  booktitle={ICML},
  year={2023}
}

Learning Rotation-Equivariant Features for Visual Correspondence

Jongmin Lee, Byungjin Kim, Seungwook Kim, Minsu Cho

The IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR) 2023 @ Vancouver

Paper Code Project Blog

@inproceedings{lee2023rotation,
  title={Learning Rotation-Equivariant Features for Visual Correspondence},
  author={Lee, Jongmin and Kim, Byungjin and Kim, Seungwook and Cho, Minsu},
  booktitle={CVPR},
  year={2023}
}

Convolutional Hough Matching Networks for Robust and Efficient Visual Correspondence

Juhong Min, Seungwook Kim, Minsu Cho

IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) 2023

Paper Code Project Blog

@article{min2023chm,
  title={Convolutional Hough Matching Networks for Robust and Efficient Visual Correspondence},
  author={Min, Juhong and Kim, Seungwook and Cho, Minsu},
  journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
  year={2023}
}

SeLCA: Self-Supervised Learning of Canonical Axis

Seungwook Kim, Yoonwoo Jeong, Chunghyun Park, Jaesik Park, Minsu Cho

Thirty-sixth Conference on Neural Information Processing Systems (NeurIPS) 2022 Workshop - Symmetry and Geometry in Neural Representations @ New Orleans

Paper Code Project Blog

@inproceedings{kim2022selca,
  title={SeLCA: Self-Supervised Learning of Canonical Axis},
  author={Kim, Seungwook and Jeong, Yoonwoo and Park, Chunghyun and Park, Jaesik and Cho, Minsu},
  booktitle={NeurIPS Workshop},
  year={2022}
}

TransforMatcher: Match-to-Match Attention for Semantic Correspondence

Seungwook Kim, Juhong Min, Minsu Cho

The IEEE / CVF Computer Vision and Pattern Recognition Conference (CVPR) 2022 @ New Orleans

Paper Code Project Blog

@inproceedings{kim2022transformatcher,
  title={TransforMatcher: Match-to-Match Attention for Semantic Correspondence},
  author={Kim, Seungwook and Min, Juhong and Cho, Minsu},
  booktitle={CVPR},
  year={2022}
}

Deep Hough Voting for Robust Global Registration

Junha Lee, Seungwook Kim, Minsu Cho, Jaesik Park

The IEEE / CVF International Conference on Computer Vision (ICCV) 2021 @ Montreal

Paper Code Project Blog

@inproceedings{lee2021deephough,
  title={Deep Hough Voting for Robust Global Registration},
  author={Lee, Junha and Kim, Seungwook and Cho, Minsu and Park, Jaesik},
  booktitle={ICCV},
  year={2021}
}

Learning to Distill Convolutional Features into Compact Local Descriptors

Jongmin Lee, Yoonwoo Jeong, Seungwook Kim, Juhong Min, Minsu Cho

IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) 2021 @ Hawaii

Paper Code Project Blog

@inproceedings{lee2021distill,
  title={Learning to Distill Convolutional Features into Compact Local Descriptors},
  author={Lee, Jongmin and Jeong, Yoonwoo and Kim, Seungwook and Min, Juhong and Cho, Minsu},
  booktitle={WACV},
  year={2021}
}

Seungwook Kim

News

Selected Publications

Improving Text-to-Image Generation with Intrinsic Self-Confidence Rewards

FreeAction: Training-Free Techniques for Enhanced Fidelity of Trajectory-to-Video Generation

RapidMV: Leveraging Spatio-Angular Representations for Efficient and Consistent Text-to-Multi-View Synthesis

Similarity-Aware Selective State-Space Modeling for Semantic Correspondence

Enhancing 3D Fidelity of Text-to-3D using Cross-View Correspondences

Learning SO(3)-Invariant Semantic Correspondence via Local Shape Transform

Efficient Semantic Matching with Hypercolumn Correlation

Stable and Consistent Prediction of 3D Characteristic Orientation via Invariant Residual Learning

TransforMatcher: Match-to-Match Attention for Semantic Correspondence

Honors & Awards

Experiences

Tech Lead / Research Scientist

PhD Intern / Bytedance Seed Team

PhD Intern / Data-Intelligent Creation-Vision and Graphics team

Technical Chair

Undergraduate Intern / 3D Map construction from LiDAR

Undergraduate Intern / Camera ISP

Undergraduate Intern / AI team

Undergraduate Intern / Data Engineering & Analysis team

Education

Integrated MSc/PhD in Graduate School of AI

BSc in Computer Sciences & Engineering

All Publications

Improving Text-to-Image Generation with Intrinsic Self-Confidence Rewards

FreeAction: Training-Free Techniques for Enhanced Fidelity of Trajectory-to-Video Generation

RapidMV: Leveraging Spatio-Angular Representations for Efficient and Consistent Text-to-Multi-View Synthesis

Similarity-Aware Selective State-Space Modeling for Semantic Correspondence

Harnessing the Power of Training-Free Techniques for Text-to-3D Generation via Score Distillation Sampling

3D Geometric Shape Assembly via Efficient Point Cloud Matching

Multi-view Image Prompted Multi-view Diffusion for Improved 3D Generation

Enhancing 3D Fidelity of Text-to-3D using Cross-View Correspondences

Learning SO(3)-Invariant Semantic Correspondence via Local Shape Transform

Efficient Semantic Matching with Hypercolumn Correlation

Stable and Consistent Prediction of 3D Characteristic Orientation via Invariant Residual Learning

Learning Rotation-Equivariant Features for Visual Correspondence

Convolutional Hough Matching Networks for Robust and Efficient Visual Correspondence

SeLCA: Self-Supervised Learning of Canonical Axis

TransforMatcher: Match-to-Match Attention for Semantic Correspondence

Deep Hough Voting for Robust Global Registration

Learning to Distill Convolutional Features into Compact Local Descriptors