alex-damian/pulse
Self-supervised image upsampling system that uses StyleGAN latent space search to generate high-resolution faces from low-resolution inputs.
Velocity · 7d
+3.6
★ / day
Trend
→steady
star history
PULSE searches the latent space of StyleGAN to find high-resolution images that downscale correctly to a given low-resolution input. The method is self-supervised and does not require paired low/high-resolution training data. It was published at CVPR 2020 and focuses on face upsampling.