Namhoon Kim
Ph.D. student in Machine Learning
Representing, reconstructing,
and understanding the visual world.
I am a Ph.D. student at Georgia Tech, advised by Sara Fridovich-Keil in the VOILA Lab. My research brings together computational imaging, signal representations, and inverse problems.
I study how representations and priors shape what we can recover from imperfect measurements. My work spans neural fields and explicit grids, training-free geometric priors for 3D reconstruction, and physics-based tomography—with an emphasis on understanding when a reconstruction can be trusted.
Research directions
What I work onRepresentations
Understanding the capacity, bandwidth, and inductive biases of neural and explicit signal representations.
02Reconstruction
Combining geometry and imaging physics to recover structure from noisy or incomplete measurements.
03Reliability
Detecting distribution shift and understanding uncertainty when learned priors guide image reconstruction.