search

NVlabs / stylegan3

Official PyTorch implementation of StyleGAN3
Other
6.3k stars 1.11k forks source link

Will the label of the dataset lead to unconditional generation? #189

Open 49xxy opened 1 year ago

49xxy commented 1 year ago

Thank you very much for your excellent work! At present, I want to use my own dataset for training, I refer to your command for training with AFHQ V2 dataset. python train.py --outdir=~/training-runs --cfg=stylegan3-t --data=~/datasets/afhqv2-512x512.zip \ --gpus=8 --batch=32 --gamma=8.2 --mirror=1 Although there are three classes of AFHQ V2 dataset, but you are trained unconditionally, so I want to know what the "dataset. Json " file in this dataset is like, is it unlabeled?As shown in the following： image

PDillis commented 1 year ago

Yes, if there are no labels, then you should only get inside the dataset.json a {"labels": null} and your model will be an unconditional one. This means you don't need to pass a class to it when generating new images/videos, but all 'classes' will be mixed, if there are any within the dataset.