Question about mean_domains, std_domains in augmentations.py

[yaelmagen]

Greetings @sophiajw and thank you for publishing this code.

I have been exploring the code you provided on GitHub and have a specific question regarding the implementation in augmentations.py.
You define the variable stats as (mean_domains, std_domains) with the following fixed values:

mean_domains = [
torch.tensor([0.3020, -2.6476, -0.9849, -0.7820, -0.2746, 0.3361, 0.1694, -1.2148]),
torch.tensor([0.1453, -1.2400, -0.9484, 0.9697, -2.0775, 0.7676, -0.5224, -0.2945]),
torch.tensor([2.1067, -1.8572, 0.0055, 1.2214, -2.9363, 2.0249, -0.4593, -0.9771]),
torch.tensor([0.8378, -2.1174, -0.6531, 0.2986, -1.3629, -0.1237, -0.3486, -1.0716]),
torch.tensor([1.6073, 1.9633, -0.3130, -1.9242, -0.9673, 2.4990, -2.2023, -1.4109]),
]

std_domains = [
torch.tensor([0.6550, 1.5427, 0.5444, 0.7254, 0.6701, 1.0214, 0.6245, 0.6886]),
torch.tensor([0.4143, 0.6543, 0.5891, 0.4592, 0.8944, 0.7046, 0.4441, 0.3668]),
torch.tensor([0.5576, 0.7634, 0.7875, 0.5220, 0.7943, 0.8918, 0.6000, 0.5018]),
torch.tensor([0.4157, 0.4104, 0.5158, 0.3498, 0.2365, 0.3612, 0.3375, 0.4214]),
torch.tensor([0.6154, 0.3440, 0.7032, 0.6220, 0.4496, 0.6488, 0.4886, 0.2989]),
]

As noted in the documentation, these values represent the (mean, std dev) of the latent space of HistAuGAN. However, I have been unable to locate the specific part of the code or documentation that describes how these means and standard deviations were computed from the data.

Could you please provide some insight or point me towards the relevant part of your project that details the process used to evaluate these fixed values?

Thank you very much for your time and assistance.