ABOUT

PEOPLE

ADMISSION

EDUCATION

ARCHIVE

BOARD

  • ARCHIVE

MV CIMF Festival

FLAIR: Frequency- and Locality-Aware Implicit Neural Representations

Sukhun Ko, Dahyeon Kye, Kyle Min, Chanho Eom, Jihyong Oh Creative Vision and Multimedia Lab

Existing implicit neural representations (INRs) are limited by spectral bias and the absence of frequency–locality awareness, which hinders their ability to model fine details. We propose FLAIR, which introduces RC-GAUSS, a raised-cosine Gaussian activation that jointly regulates frequency selection and spatial localization through learnable parameters under the time–frequency uncertainty principle (TFUP), along with a complementary WEGE module that leverages discrete wavelet transform (DWT) to compute energy scores and guide frequency information. FLAIR outperforms existing approaches in both 2D image fitting and restoration and further generalizes to 3D representation tasks.