301-1118, Seoul National University, 1 Gwanak-ro, Gwanak-gu, Seoul 08826 📱*+82 10 4230 0574 |* ☎ +82 2 880 9570 | 📧 [email protected] | 💻 Google Scholar

Education

• Integrated Ph. D. program, Seoul National University in Department of Electrical and Computer Engineering

  Supervised by Prof. Byoungho Lee ([email protected], webpage) Supervised by Prof. Yoonchan Jeong ([email protected], webpage)

2019 - Present

• Bachelor of Physics, Ewha Womans University in College of Natural Science

2014 - 2019

Research Interests

• Holographic display & Computer generated hologram
• Machine learning / Deep learning
• Computational imaging
• Computer vision
• Augmented reality (AR) & Virtual reality (VR)

Technical skills

• Optical setups: Experiences on laser experiments, such as holographic display, digital holography, holographic printer
  1. Holographic display setup: Used in the article “Deep neural network for multi-depth hologram generation and its training strategy”
  2. Digital holography: Used in the project “Research and development of overlapped fingerprint identification technology using laser”
• Programming and tools: Python, Tensorflow, Pytorch, MATLAB, C++
  1. Python, Pytorch: Used in the article “FinSNet: end-to-end separation of overlapped fingerprints using deep learning” and multiple project
  2. Python, Tensorflow: Used in the article “Deep neural network for multi-depth hologram generation and its training strategy”
  3. MATLAB: Used in the conference proceeding “Complex hologram generation of multi-depth images using deep neural network”

Research Experience - Topics

• Deep learning technology and computer-generated holograms

  Computer-generated holograms (CGH) are traditionally acquired using numerical propagation models, which require substantial computational resources. By incorporating deep learning, the speed of CGH computation can be greatly improved. Neural networks are trained to learn optical propagation models for generating multi-depth holograms, using artificial intensity maps and target holograms. This research is detailed in the paper, “Deep neural network for multi-depth hologram generation and its training strategy,” published in Optics Express (2020).

• Hologram conversion for various holographic displays

  Numerical propagation models for hologram generation require specific parameters from the holographic display setup, such as the wavelength of the light source and the pixel pitch of the hardware. Since each holographic display has unique parameters, the hologram must be regenerated for each system. Utilizing a deep neural network can significantly reduce the computational cost associated with generating holograms for the same content. This research is detailed in the paper, “Fast deep-trained transformation technique for computer-generated holograms,” presented at SPIE Photonics West (2023).