Does anyone know what are the mean and std for LSUN dataset.
It would have been nice if the data loaders in pytorch had these values at time of construction of the data loader.
For instance if the dataloader would have the following args
normalize loaded by default.
data = tv.datasets.LSUN(....) loader = torch.utils.data.DataLoader(data, normalize=True) loader.mean <--- print mean for dataset loader.std <-- print std for dataset
normalize=True would normalize the dataset by default no need to add it to transforms unless the user requires it by manually setting
normalize=False, then s/he could pass whatever normalization they would like via the transforms.
Just a thought?