search

MIC-DKFZ / batchgenerators

A framework for data augmentation for 2D and 3D image classification and segmentation
Apache License 2.0
1.09k stars 221 forks source link

not all Transforms have a p_per_sample parameter #73

Open filippocastelli opened 3 years ago

filippocastelli commented 3 years ago

I noticed that not all the Transforms have a p_per_sample probability parameter, is this an intended behavior for some reason? ex: Rot90Transform has it, MIrrortransform has not.

Also noticed that this was already addressed in pull request #66

FabianIsensee commented 3 years ago

Hi, I will look into this. Best, Fabian

FabianIsensee commented 3 years ago

Hi, can you please list the Transforms for which you would like to have a p_per_sample? I was under the impression that if we are doing mirroring we might as well apply it uniformly to all samples. But i can see that other users would like to have more freedom. I will add it to MirrorTransform now and can also add it to other transforms of your choice. Best, Fabian