This is an implementation of the paper Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains using TensorFlow 2.3.
I also added a learnable parameter instead of using a fixed random Gaussian parameter. The PSNR curve over optimization steps(x25):
The output images of each method:
