Skip to content

Latest commit

 

History

History
27 lines (16 loc) · 787 Bytes

README.md

File metadata and controls

27 lines (16 loc) · 787 Bytes

Data preparation for MetFaces experiments

The folder MetFaces/prepare_data contains the code to prepare data for the MetFaces experiments.

Pre-trained FFHQ image classifier

Since we apply the FFHQ image classifier to annotate the MetFaces data, you first need to follow the instruction Training FFHQ image classifier to get the pre-trained FFHQ image classifier.

Data generation

Generate 10k images and latent variables of StyleGAN2-ADA (including w and z):

bash scripts/run_gen_batch.sh

Use the pre-trained FFHQ image classifier to annotate the generated images:

bash scripts/run_metfaces_labeling.sh

The resulting pairs of latent variables and labels will be used to train latent classifiers.