Seong-heum Kim

Senior Research Engineer [CV]
Intelligent Image Processing Research Center
Korea Electronics Technology Institute (KETI)

Contact: +82-31-789-7469
E-mail: [at] (or

Address: Global R&D Center, Pangyo Techno Valley
22, Daewangpangyo-ro 712beon-gil, Bundang-gu, Seongnam-si, Gyeonggi-do, Republic of Korea

Currently, I work for Computer Vision Team at KETI. For the past seven years, I had been trained at Robotics and ComputerVision (RCV) Lab. by my academic advisor, Prof. In So Kweon. For my alternative military service 2011~2014, I had been a junior research engineer at Digital TV (DTV) SoC Dept., System IC (SIC) Center, LG Electronics. I received my B.S. and M.S./Ph.D. in Electrical Engineering from Yonsei University in 2007 and from KAIST in 2010, 2018 respectively. My research interests include computer vision/graphics, deep learning, and industrial applications of artificial intelligence.



Under revision

Two Phase Approach for Instance-level 3D Model Retrieval

Seong-heum Kim, Youngbae Hwang, In So Kweon
Image and Vision Computing (IVC), To appear

RANUS: RGB and NIR Urban Scene Dataset for Deep Scene Parsing

Gyeongmin Choe, Seong-heum Kim, Sunghoon Im, Joon-Young Lee, Srinivasa G. Narasimhan, In So Kweon
IEEE Robotics and Automation Letters (RAL), 3(3), pp. 1808-1815, July 2018
[project page] [pdf] [bibtex]

Deep Representation of Industrial Components using Simulated Images

Seong-heum Kim, Gyeongmin Choe, Byungtae Ahn, In So Kweon
In Proc. of IEEE International Conference on Robotics and Automation (ICRA), May. 2017
[pdf] [data] [bibtex]

Category-specific Salient View Selection via Deep Convolutional Neural Networks

Seong-heum Kim, Yu-Wing Tai, Joon-Young Lee, Jaesik Park, In So Kweon
Computer Graphics Forum (CGF), Invited to Eurographics 2017, Apr. 2017
[pdf] [2D data] [3D data] [bibtex]

2D-3D Pose Estimation using Multi-view Object Co-segmentation

Seong-heum Kim, Yunsu Bok, In So Kweon
Journal of Korea Robotics Society (KROS), Best Paper Award, 12(1), pp. 33-41, Feb. 2017
[pdf] [bibtex]

Multi-view Object Extraction with Fractional Boundaries

Seong-heum Kim, Yu-Wing Tai, Jaesik Park, In So Kweon
IEEE Transactions on Image Processing (TIP), 25(8), pp. 3639-3654, Aug. 2016
[pdf] [supple] [video] [bibtex]

Two-phase Approach for Multi-view Object Extraction

Seong-heum Kim, Yu-Wing Tai, Yunsu Bok, Hyeongwoo Kim, In So Kweon,
In Proc. of IEEE International Conference on Image Processing (ICIP), Sep. 2011
[pdf] [bibtex]

Mixing Collaborative and Hybrid Vision Devices for Robotic Applications

Jean-Charles BazinSeong-heum Kim, Dong-Geol Choi, Joon-Young Lee, In So Kweon
Journal of Korea Robotics Society (KROS), Sep. 2011
[pdf] [bibtex]



  • 김성흠유한나, 박한무, 황영배골격의 길이 정보를 이용한 제스쳐 인식 방법 및 처리 시스템 (GESTURE RECOGNITION METHOD AND PROCESSING SYSTEM USING SKELETON LENGTH INFORMATION), 국내출원 2018-0165196 (2018.12.19)
  • 김성흠박민규, 황영배스테레오 카메라와 근적외선 조명/센서 융합 기반 임의 객체 길이 측정 및 시스템 (OBJECT LENGTH MEASUREMENT METHOD AND SYSTEM BASED ON STEREO CAMERA AND NEAR-INFRARED ILLUMINATION/SENSOR), 국내출원 2018-0165170 (2018.12.19)
  • 김성흠김정호, 황영배심층 구조 학습 기반 사람 계수 방법 및 시스템 (APPARATUS AND METHOD FOR COUNTING PEOPLE USING DEEP LEARNING), 국내출원 2018-0079358 (2018.07.09)
  • 권인소, 김성흠. 삼차원 모델의 대표 이미지 생성 장치 방법 (APPARATUS AND METHOD FOR GENERATING SALIENT IMAGES OF 3D MODELS), 국내등록 1018318210000 (2018.02.19)
  • 권인소, 김성흠. 가분적 외곽선을 가지는 다시점 객체 추출 방법 장치(METHOD AND APPARATUS FOR EXTRACTING MULTI-VIEW OBJECT WITH FRACTIONAL BOUNDARIES), 국내 등록 1018126640000 (2017.12.20)


Honors & Awards

  • 우수신진연구자상(ICROS19-135호), 제어로봇시스템학회 2019
  • 제28회 과학기술우수논문상(과총 제2018-061호), 한국과학기술단체총연합회(KOFST) 2018
  • Best Paper Award, “2D-3D Pose Estimation using Multi-view Object Co-segmentation,” Journal of Korea Robotics Society (KROS) 2017
  • Gold prize, “Eye-contact by Embedding a Camera into a Transparent Display Panel,” LG CTO Idea Contest 2011
  • High Honors at Early Graduation, Yonsei University, Aug. 2007