Exploring the potential of StyleGAN projection for quantitative maps from diffusion-weighted MR images