Fernando Pérez-García

Hi, @justusschock. I agree it makes sense to make any methods not using `self` static.

Sorry I haven't reviewed this earlier. Could you please verify that #981 solves your issue?

> If it's included in the RandomFlip, you would have to specify which LabelMap the remapping should be applied to as you mentioned Actually, shouldn't there already be a kwarg...

Maybe the label map could optionally take a path to a CSV file [like this](https://raw.githubusercontent.com/fepegar/highresnet/master/GIFNiftyNet.ctbl). The laterality would ideally be inferred from the path, so the format of the table...

Interestingly, I've been trying to use the above approach (Flip + Remap) on the brain parcellation notebook on Colab, but didn't work: ```python num_augmentations = 10 results = [] import...

> Yes, since it's compatible with the include and exclude system from Transform. Ah, sure. Do you mean there's no need for a new kwarg because we can already use...

Hi, @dmus. I haven't fully understood how this works, but it certainly looks promising. So would the samplers need to be reimplemented as well? How random is this? The way...

Nice. Using a map-style dataset would ensure that patches are extracted from each subject once, and would let us define an epoch, right? I'm not sure why that approach would...

The buffered shuffle dataset seems to be available in the nightly version: ``` Py 3.8.5 (test)  ~  pip install numpy  ✔  26% (2:18)  24.63s ...

Hi, @dmus. This looks interesting. It's nice to be able to leverage newer PyTorch classes. I have some questions: 1. Using multiple datasets to sample one randomly seems like a...