How content-conditioned style encoder is implemented?

NVlabs / imaginaire

NVIDIA's Deep Imagination Team's PyTorch Library

Other

3.99k stars 444 forks source link

How content-conditioned style encoder is implemented? #144

Closed lichong952012 closed 2 years ago

lichong952012 commented 2 years ago

How the elementwise product is implemented between style encoder and content encoder in COCO-FUNIT? It looks the same as FUNIT.

mingyuliutw commented 2 years ago

In FUNIT, the style code depends on the style image only. The upper branch of the figure is absent. Of course, we do not use constant style bias in FUNIT.