Minho Choi
Ulsan National Institute of Science and Technology · 工学
研究室紹介
Professor Minho Choi's research lab specializes in nanophotonics and metasurface engineering, focusing on the design and optimization of flat optical components with tailored wavefront manipulation capabilities. The lab pioneers differentiable optical design methods using machine learning frameworks like TensorFlow, enabling end-to-end optimization of RGB metasurfaces for custom point spread functions. A key research direction involves bridging the sim-to-real gap in meta-optical systems through knowledge distillation, ensuring robust performance across simulation and real-world implementations. The lab also explores fundamental photonic phenomena such as flatbands and angle-insensitive resonances in 2D photonic lattices for enhanced light-matter interactions.
Research Overview
Research Output Trend
Figures are computed from collected data and may differ slightly.
Selected Papers
6Photonic flatbands offer promising light-matter interaction due to their unique slow-light nature. In recent years, flatbands have also attracted significant interest in optical engineering because of their angle-insensitive resonant characteristics. However, to date, no studies have reported the dispersionless behavior of flatbands under arbitrary two-dimensional incident angles and polarizations. Here, we present a two-dimensional photonic flatband created using a silicon metasurface with a Li
Includes a Colab notebook for designing RGB metasurfaces that generate custom point spread functions using TensorFlow. Implements differentiable phase modeling, band-limited angular spectrum method, and loss minimization across RGB wavelengths.
Knowledge distillation for minimizing the sim-to-real gap of the PSF engineered meta-optical encoder
Knowledge distillation for minimizing the sim-to-real gap of the PSF engineered meta-optical encoder
Includes a Colab notebook for designing RGB metasurfaces that generate custom point spread functions using TensorFlow. Implements differentiable phase modeling, band-limited angular spectrum method, and loss minimization across RGB wavelengths.