This is a simple PyTorch implementation of Fourier Feature Networks, translated from the officially released code.
For more information, please refer to:
Tancik et al. Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains, NeurIPS 2020. (Paper|Project)
This code takes image coordinate (x,y) as input, and regresses/outputs the corresponding RGB pixel values.
- none: Standard linear regression network (no mapping of input coordinate (x,y)).
- basic: Basic Fourier feature mappings of (x,y).
- gauss_1: Gaussian Fourier feature mappings of (x,y) with scale=1.
- gauss_10: Gaussian Fourier feature mappings of (x,y) with scale=10.
- gauss_100: Gaussian Fourier feature mappings of (x,y) with scale=100.
| Original image | none | basic |
|---|---|---|
 |
 |
 |
| gauss_1 | gauss_10 | gauss_100 |
 |
 |
 |
