RameenAbdal / CLIP2StyleGAN

CLIP2StyleGAN: Unsupervised Extraction of StyleGAN Edit Directions (SIGGRAPH 2022)

RameenAbdal/CLIP2StyleGAN Issues

RuntimeError: Error(s) in loading state_dict for Generator: Unexpected key(s) in state_dict: "conv1.activate.bias", "convs.0.activate.bias", "convs.1.activate.bias", "convs.2.activate.bias", "convs.3.activate.bias", "convs.4.activate.bias", "convs.5.activate.bias", "convs.6.activate.bias", "convs.7.activate.bias", "convs.8.activate.bias", "convs.9.activate.bias", "convs.10.activate.bias", "convs.11.activate.bias", "convs.12.activate.bias", "convs.13.activate.bias".
Updated 6 months ago
Would those extracted directions be open-source?
Updated 2 years ago
is it reasonable to feed the net a soft combination of the 49408 embeddings?
Updated 2 years ago
How can I make the latents in data_latents and latents_ in edits_sg?
Updated 2 years ago
Custom Direction
Updated 2 years ago
How can i make extracted_features by myself?
Closed 2 years ago1
Where are the words of predictions in optimize_dir.py coming from?
Updated 2 years ago
The codes for image editing are missing
Closed 2 years ago1