CV

M.S. in The Cho Chun Shik Graduate School of Green Transportation

Korea Advanced Institute of Science and Technology(KAIST), Republic of Korea

• Thesis | Vision-based Real-time Welding Line Detection Algorithms for Automatic Welding Robot [3]
• Advisor | Kyung-soo Kim

B.S. in Electronic Engineering

Inha University, Republic of Korea

• First honor of fall graduation in Department of Electronic Engineering

NAVERZ, AI Researcher

NAVERZ is one of the most popular company in the realms of metaverse service since 2019, having 300 millions users, who are distributed mainly in south-east Asia and North America. By acquiring the AI team of PlaceA, NAVERZ teamed up the motion AI team, whose primary role has creating and researching the interface between the real world and metaverse service of Zepeto—I mainly charge for addressing multiple problems by adapting a multi-view camera to enhance a joint estimation result and to cover all invisible and occlusive area.

• Responsibilities included the following
  • Designing and interpreting 2D joints model to enhance model quality(inference speed and accuracy)
  • Developing 3D human joints estimation model to compensate for dead-zone via multi-view cameras
  • Establishing and evaluating synthetic data to accelerate joints estimation results in dynamic movement

PLACEA, AI Researcher

PLACE A, an AI-tech startup in Korea, mainly developed image-based scaleable AI solutions that are used in services related to human motions. I joined as a Researcher, tasked with studying how to estimate human pose, including 2D, 3D joint coordinates, and mesh level using deep learning. PLACE A provided accurate, real-time, and intuitive joint information to the users, with the team’s model playing a pivotal role as a significant feature in one of the popular metaverse services. Moreover, I also participated in the research team on the interpretive learning model for authenticity verification of whether a specific product is authentic or not.

• Responsibilities included the following
  • Re-designed and optimized the real-time monocular 3D joint estimation model to boost model inference speed and accuracy.
  • Classified authentication for examination of used goods in the online platform using uncertainty-based AI model
• Results
  • Contributed to achieve 160FPS for 3D pose estimation inference speed, on C++ binary(GTX 3060)
  • Increased to 30% pose accuracy(MPJPE) in dynamic pose situation(e.g. yoga, exercise)
  • Established anomaly detection model with having an accuracy of 99%

TMAX, Researcher

Tmax is the domestic hidden champion who leads the field of system software sectors such as database and middleware, awarded by providing its product to Hyundai and multiple domestic bank companies. I joined as a Researcher in the 2D Graphics team, developed the C++-based 2D graphic library for the logic of drawing and rendering objects and fonts as well as the conventional algorithms of image processing on TmaxOS and relevant. I also conducted deep learning-based computing vision algorithms for various software products.

• Responsibilities included the following
  • Improved graphic rendering algorithms within TmaxOS, a Linux-based operating system, for a user experience equivalent to Windows graphics.
  • Designed a virtual background based on deep learning for HyperMeeting, a web-based video-conferencing service
• Results
  • Built a prototype for deep-learning-based virtual background model for HyperMeeting
  • Contributed to release TmaxOffice in Windows to enhance font engine
  • Contributed to launch HyperMeeting service's front-end development

Hyundai Heavy Industry, Researcher

Hyundai Heavy Industry is the global leading shipbuilding company. I was recruited as a Researcher to develop various automation robots to enhance plant productivity. Mainly I contributed to developing an LNG tank welding robot that was part of the most essential and sophisticated process throughout the LNG cargo shipbuilding process. HHI enabled to acquire the certification from shipowners and classification society, which are necessary for the deployment of the robot in the factories.

• Responsibilities included the following
  • Developed a vision-based algorithm for tracking the welding line in the process.
  • Developed an automatic welding robot for the gas tanks of LNG carriers fitted with an automatic welding line tracking robot.
• Results
  • Revised and reprogrammed whole electronics architecture on welding robots including replacement a type of main processor
  • Acquired the certification from the shipowners and the classification society, necessary for the deployment of the robot into the factories

Korea Advanced Institute of Science and Technology(KAIST), Graduate Research Assistant

• Conducted projects
  • Development welding line tracking vision-based algorithm of LNG cargo welding robot sponsored by Hyundai Heavy Industry
  • Development vehicle body velocity sensor using Modulated Motion Blur
  • Development automatic parking algorithm using single CCTV in scaled down environment

• National Student Scholarship for Master's Program at KAIST

• Hanjin Group Scholarship

• Second place in Campus Start-ups Competitions

• Academic Excellence Scholarship for top 3% of the department

• Samsung Dream Class
• Supporting low-income students in the middle school to provide math lectures