About me

I’m Youngseok Kim (김영석), currently a computer vision research engineer at 42dot. I am developing a perception model for autonomous driving system and I am in charge of 3D object detection using camera and radar.

I obtained my PhD degree from KAIST in 2023, advised by Prof. Dongsuk Kum and I graduated with KAIST College of Engineering PhD Dissertation Award. My research interest lies in 3D computer vision and its application to autonomous driving. Specifically, I am focusing on developing a robust and cost-effective 3D perception system using camera and radar, self-supervised and active learning to make the cost- and time-consuming training process efficient.

Publications

• LabelDistill: Label-guided Cross-modal Knowledge Distillation for Camera-based 3D Object Detection
  Sanmin Kim, Youngseok Kim, Sihwan Hwang, Hyeonjun Jeong, Dongsuk Kum
  European Conference on Computer Vision (ECCV), 2024
  [Paper] [Code]

• CRN: Camera Radar Net for Accurate, Robust, Efficient 3D Perception
  Youngseok Kim, Juyeb Shin, Sanmin Kim, In-jae Lee, Jun Won Choi, Dongsuk Kum
  IEEE/CVF International Conference on Computer Vision (ICCV), 2023
  [Paper] [Code] [Video Demo] [Poster] [Presentation]
   -   Preliminary version is published in ICLR SR4AD Workshop [Paper] [Presentation]
   -   Presented at 2023 KAIST Future Mobility Conference [Presentation (Korean)]
  Ranked 1st among camera-radar methods on nuScenes as of March 2023
  

• Predict to Detect: Prediction-guided 3D Object Detection using Sequential Images
  Sanmin Kim, Youngseok Kim, In-Jae Lee, Dongsuk Kum
  IEEE/CVF International Conference on Computer Vision (ICCV), 2023
  [Paper] [Code]

• CRAFT: Camera-Radar 3D Object Detection with Spatio-Contextual Fusion Transformer
  Youngseok Kim, Sanmin Kim, Jun Won Choi, Dongsuk Kum
  AAAI Conference on Artificial Intelligence (AAAI), 2023
  [Paper] [Video Demo] [Poster] [Presentation]
   -   Introduced in 2023 KAIST Annual R&D Report [Link (English)] [Link (Korean)]
   -   Ranked 1st among camera-radar methods on nuScenes as of July 2022

• Joint Semi-Supervised and Active Learning via 3D Consistency for 3D Object Detection
  Sihwan Hwang, Sanmin Kim, Youngseok Kim, Dongsuk Kum
  IEEE/RSJ International Conference on Robotics and Automation (ICRA), 2023
  [Paper] [Video Demo]
  

• Boosting Monocular 3D Object Detection with Object-Centric Auxiliary Depth Supervision
  Youngseok Kim, Sanmin Kim, Sangmin Sim, Jun Won Choi, Dongsuk Kum
  IEEE Transactions on Intelligent Transportation Systems (T-ITS), 2022 (Impact Factor 9.551)
  [Paper] [Video Demo]
   -   Ranked 1st/3rd among KITTI monocular BEV/3D detection as of 2021 April
  Introduced in 2021 KAIST breakthroughs [Link]
  

• Sequential Image-based 3D Object Detection with Location Refinement
  Sangmin Sim, Youngseok Kim, Dongsuk Kum
  IEEE International Conference on Pattern Recognition (ICPR), 2022
  [Paper]

• GRIF Net: Gated Region of Interest Fusion Network for Robust 3D Object Detection from Radar Point Cloud and Monocular Image
  Youngseok Kim, Jun Won Choi, Dongsuk Kum
  IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020
  [Paper] [Video Demo] [Presentation]
   -   Presented at Seoul Mobility Show 2021 Conference

• Low-level Sensor Fusion Network for 3D Vehicle Detection using Radar Range-Azimuth Heatmap and Monocular Image
  Jinhyeong Kim*, Youngseok Kim*, Dongsuk Kum
  Asian Conference on Computer Vision (ACCV), 2020
  [Paper] [Presentation]
  

• Deep Learning based Vehicle Position and Orientation Estimation via Inverse Perspective Mapping Image
  Youngseok Kim, Dongsuk Kum
  IEEE Intelligent Vehicles Symposium (IV), 2019
  [Paper] [Video Demo] [Poster]
  Oral presentation, 5.8% acceptance rate for oral-presentation papers

Other Publications

• Autonomous Driving Technology Trend and Future Outlook: Powered by Artificial Intelligence
  Sanmin Kim, Youngseok Kim, Hyeongseok Jeon, Dongsuk Kum, Kibeom Lee
  Transactions of Korean Society of Automotive Engineers (Trans. KSAE), 2022
  

• Vehicle Distance Estimation using Convolutional Neural Network on Inverse Perspective Mapping Image
  Youngseok Kim, Dongsuk Kum
  Korean Society of Automotive Engineers Annual Conference (KSAE), 2019
  

• Fault Tolerant Vehicle Detection Using Camera-LiDAR Sensor Fusion: Multi-channel Faster R-CNN
  Youngseok Kim, Dongsuk Kum
  Korean Society of Automotive Engineers Annual Conference (KSAE), 2018
  

Patents

• Electronic Device for Perceiving Three-Dimension Environment based on Camera and Radar, and Operating Method Thereof
  Dongsuk Kum, Youngseok Kim
  US, Application Number: 18/502,678
  KR, Application Number: 10-2023-0096379
  

• Electronic Device for Camera and Radar Sensor Fusion-based Three-dimensional Object Detection and Operating Method Thereof
  Dongsuk Kum, Youngseok Kim
  US, Registration Number: 11,754,701 [Patent]
  DE, PCT, Application Number: 10 2021 106 518.6, PCT/KR2021/002916
  KR, Registration Number: 10-2168753-0000
  

• Simultaneous Traffic Participants Detection and Localization via Bird’s Eye View Image
  Dongsuk Kum, Youngseok Kim
  KR, Registration Number: 10-2003387-0000
  

• Distance Measuring Device Using Mono Infrared Camera and Method Thereof
  Dongsuk Kum, Youngseok Kim, Seoung Jun Lee
  KR, Registration Number: 10-1918887-0000
  

Professional Services

Conference Reviewer for AAAI, CVPR, ECCV, ICRA, IROS, ITSC, IV
Journal Reviewer for IJAT, NeuroComputing, RA-L, T-ITS, T-IV, TPAMI