search

POSTECH-CVLab / PyTorch-StudioGAN

StudioGAN is a Pytorch library providing implementations of representative Generative Adversarial Networks (GANs) for conditional/unconditional image generation.
https://github.com/MINGUKKANG
Other
3.42k stars 341 forks source link

TSNE Analysis, K-Nearest Neighbor Analysis, Frequency Analysis and Semantic Factorization #186

Open davidHemf opened 1 year ago

davidHemf commented 1 year ago

Hi & good day!

Thanks for the paper, the repo and codes. I was searching through folders to find out the source code for image analysis, but didn't see anything relate to KNN, T-SNE, Frequency and semantic factorization. Where can I find them? And is it possible to run any of analysis with just one CUDA? CUDA_VISIBLE_DEVICES=0 instead of CUDA_VISIBLE_DEVICES=0,...,N

In the README, it mention to "3 differentiable augmentations" so how to activate or deactivate augmentation during training?

Thanks again!

mingukkang commented 1 year ago

Hi & good day!

Thanks for the paper, the repo and codes. I was searching through folders to find out the source code for image analysis, but didn't see anything relate to KNN, T-SNE, Frequency and semantic factorization. Where can I find them?

Here :)

And is it possible to run any of analysis with just one CUDA? CUDA_VISIBLE_DEVICES=0 instead of CUDA_VISIBLE_DEVICES=0,...,N

Yes, you can.

In the README, it mention to "3 differentiable augmentations" so how to activate or deactivate augmentation during training?

You can activate or deactivate using our configuration system. The previous hyperlink shows our options for controlling differentiable augmentations. You can turn on or off the differentiable operations by changing .yaml file. Please refer to BigGAN-DiffAug.yaml and StyleGAN-ADA.yaml yaml primitives.

davidHemf commented 1 year ago

Hi again @mingukkang Thanks for the great work :) and wonderful reply!