Junghoon Seo

Hi, I was the technical leader of the research center at SI Analytics, a satellite/aerial imagery analysis company in South Korea. At the same time, I also study for a Master’s degree at KAIST HCI Tech Lab (Prof. Sangho Yoon). My main scope of work is directing research projects on computer vision and machine learning for remote sensing. Besides, I am interested in computer graphics, parallel computing, and intelligent sensing technique for human-computer interaction. Don’t hesitate to contact me via Email or LinkedIn if you have any business!

Specialties

• Machine Learning, Computer Vision, Computer Graphics, and their Applications to Remote Sensing
• Parallel Computing on HPC and General-Purpose Computing on GPU
• Intelligent Sensing Techniques for Human-Computer Interaction
• Process Management and Policymaking for AI-driven Product-Business Alignment

Career

• Sep 2020 - Sep 2024, Technical Leader of Research Center & Co-founder @ SI Analytics
• Jul 2017 - Feb 2020, ML/CV Research Scientist @ Satrec Initiative

Education

• Feb 2023 - Current, KAIST, Daejeon, South Korea
  • Studying for Master’s degree in Graduate School of Culture Technology
• Mar 2014 - Feb 2021, GIST, Gwangju, South Korea
  • B.S degree, Major in Electrical Engineering and Computer Science

Publications (Google Scholar)

• Posture-Informed Muscular Force Learning for Robust Hand Pressure Estimation
  • Junghoon Seo, Kyungjin Seo, Hanseok Jeong, Sangpil Kim, Sang Ho Yoon
  • NeurIPS. 2024. To be Appeared.
• A Billion-scale Foundation Model for Remote Sensing Images
  • Keumgang Cha, Junghoon Seo, and Taekyung Lee
  • IEEE J-STARS. 2024. Link
• Self-Pair: Synthesizing Changes from Single Source for Object Change Detection in Remote Sensing Imagery
  • Minseok Seo, Hakjin Lee, Yongjin Jeon, and Junghoon Seo
  • IEEE WACV. 2023. Link
• Geometric Remove-and-Retrain (GOAR): Coordinate-Invariant eXplainable AI Assessment
  • Yonghyun Park, Junghoon Seo, Bomseok Park, Seongsu Lee, and Junghyo Jo
  • NeurIPS Workshop. 2023.
• Prototype-oriented Unsupervised Change Detection for Disaster Management
  • Youngtack Oh, Minseok Seo, Kim Doyi, and Junghoon Seo
  • NeurIPS Workshop. 2023. Link
• Semi-Implicit Hybrid Gradient Methods with Application to Adversarial Robustness
  • Beomsu Kim and Junghoon Seo
  • AISTATS. 2022. Link Video
• Quantile Autoencoder with Abnormality Accumulation for Anomaly Detection of Multi-variate Sensor Data
  • Seunghyoung Ryu, Jiyeon Yim, Junghoon Seo, Yonggyun Yu, and Hogeon Seo
  • IEEE Access. 2022. Link
• Contrastive Multiview Coding With Electro-Optics for SAR Semantic Segmentation
  • Keumgang Cha, Junghoon Seo, and Yeji Choi
  • IEEE GRSL. 2021. Link
• Training Domain-invariant Object Detector Faster with Feature Replay and Slow Learner
  • Chaehyeon Lee, Junghoon Seo, and Heechul Jung
  • CVPR Workshop. 2021. Link
• On the Power of Deep but Naive Partial Label Learning
  • Junghoon Seo and Joon Suk Huh
  • IEEE ICASSP. 2021. Link
• NL-LinkNet: Toward Lighter but More Accurate Road Extraction with Non-Local Operations
  • Yooseung Wang, Junghoon Seo, and Taegyun Jeon
  • IEEE GRSL. 2021. Link
• Revisiting Classical Bagging with Modern Transfer Learning for On-the-fly Disaster Damage Detector
  • Junghoon Seo, Seungwon Lee, Beomsu Kim, and Taegyun Jeon
  • NeurIPS Workshop. 2019. Link
• Deep Closed-Form Subspace Clustering
  • Junghoon Seo, Jamyoung Koo, and Taegyun Jeon
  • ICCV Workshop. 2019. Link
• Why are Saliency Maps Noisy? Cause of and Solution to Noisy Saliency Maps
  • Beomsu Kim, Junghoon Seo, Jeongyeol Choe, Jamyoung Koo, Seunghyeon Jeon, and Taegyun Jeon
  • ICCV Workshop. 2019. Link
• Bridging Adversarial Robustness and Gradient Interpretability
  • Beomsu Kim, Junghoon Seo, and Taegyun Jeon
  • ICLR Workshop. 2019. Link
• RBox-CNN: Rotated Bounding Box based CNN for Ship Detection in Remote Sensing Image
  • Jamyoung Koo, Junghoon Seo, Seunghyun Jeon, Jeongyeol Choe, and Taegyun Jeon
  • ACM SIGSPATIAL. 2018. Link
• Noise-adding Methods of Saliency Map as Series of Higher Order Partial Derivative
  • Junghoon Seo, Jeongyeol Choe, Jamyoung Koo, SeungHyun Jeon, Beomsu Kim, and Taegyun Jeon
  • ICML Workshop. 2018. Link
• Domain Adaptive Generation of Aircraft on Satellite Imagery via Simulated and Unsupervised Learning
  • Junghoon Seo, Seunghyun Jeon, and Taegyun Jeon
  • ACML Workshop. 2017. Link
• Multi-task Learning for Fine-grained Visual Classification of Aircraft
  • Seunghyun Jeon, Junghoon Seo and Taegyun Jeon
  • ACML Workshop. 2017.

Under Review / On-going Scripts (Working Titles)

• Pressure Estimation in Hand-based Interaction (2 Projects)
• Pitfalls and Fixes in Attribution-based Explainer Evaluation (2 Projects)
• Improved Off-policy Evaluation from Multiple Logging Policies
• Improved Query-based Object Detection

Awards

• 2024, my lab’s demo, which included the project I was leading, had been selected as the Popular Choice Winner in CHI 2024.
• 2021, I graduated the college with Best Undergraduate Thesis Award.
• 2020, my team ranked the 5th place in xView2 Challange.
• 2018, my team ranked the 3rd-4th place in CVPR NTIRE Super-resolution Challange.
• 2018, my team ranked the 2nd place in DOTA Challange.
• 2016, my team got the grand prize at KISTI National Supercomputing Competition.
• 2016, I got the Qualcomm-GIST Innovation Award.

Talks

• Query as Representation: A Paradigm Shift in Computer Vision Caused by DETR @ Online. Nov 2022. YouTube (Korean)
• Recent XAI Trends in Deep Learning Era: (Under-)specification and Approaches @ KAERI. Feb 2020.
• Back to the Representation Learning with focusing on Visual Self-supervision @ ETRI. Jul 2019. Presentation (Korean)
• Deep Perceptual Super-resolution: Going Beyond Distortion @ KARI. Jul 2018. Presentation

Review-serving

• IEEE Transactions on Pattern Analysis and Machine Intelligence
• IEEE Transactions on Neural Networks and Learning Systems
• IEEE Geoscience and Remote Sensing Letters
• Several Conferences including NeurIPS, CVPR, AISTATS, BMVC…
• Several Workshops on ICCV, ICLR, CVPR…

Patents

• Method for detecting object, granted, 10-2364882
• Method for detecting on-the-fly disaster damage based on image, granted, 10-2255998
• Method for predicting frame using deep learning, application, 10-2023-0115699
• Method, system, and computing device for generating an alternative image of a satellite or aerial image in which an image of an area of interest is replaced, application, 10-2023-0075347
• Method of training object prediction models using ambiguous labels, application, 10-2022-0000169
• Method for detecting object, application, 10-2021-0168545
• Method to detect object, application, 10-2021-0135451
• Method for data clustering, application, 10-2020-0107206

Note

Due to confidentiality provisions related to the field, I cannot disclose the outline of the project to the public.