search

koushiksrivats / face_attribute_attack

Official implementation of the paper "Evading Forensic Classifiers with Attribute-Conditioned Adversarial Faces" (CVPR 23)
https://koushiksrivats.github.io/face_attribute_attack/
40 stars 5 forks source link

Dataset Split as Train/Test and Real Fake #6

Closed wasim004 closed 3 months ago

wasim004 commented 3 months ago

Hi, Thank you for the code you provided！

I am a bit confuse about how to split the dataset in Train/Test and also which data among two separate directories including (stylegan2-config-f-psi and images_1024*1024) to be placed in real and fake directories? I could not find any explanation regarding splitting and placing the data of the aforementioned two directories that contains the data.

To be precise: I have two directories now like:

Fake_Attrib_Attack/images_1024*1024 (contains 69 subfolders/subfolder/1000 images) Fake_Attrib_Attack/stylegan2-config-f-psi(contains 100 subfolders/subfolder/1000 images)

1- How to split these two directories in Train/Test? 2- How and which data to be placed in (Train/Real, Train/Fake) and in (Test/Real, Test/Fake)?

Thank You!

wasim004 commented 3 months ago

Hi,

I find the details in the manuscript. Thank YoU!