Fixes #98 - Enable the use of `dataset_volume.py` for unsupervised cases (label/mask = None) - Githubissues

zudi-lin / pytorch_connectomics

PyTorch Connectomics: segmentation toolbox for EM connectomics

http://connectomics.readthedocs.io/

MIT License

172 stars 77 forks source link

Fixes #98 - Enable the use of `dataset_volume.py` for unsupervised cases (label/mask = None) #101

Closed Lauenburg closed 2 years ago

Lauenburg commented 2 years ago

I added multiple switches to dataset_volume.py that ensure its functionality in the case that labels and/or valid_mask is set to None.

I tested my adaptation with dummy data for the following cases:

mode='train', volume=data, lable=data, valid_mask=data
mode='train', volume=data, lable=None, valid_mask=data
mode='train', volume=data, lable=data, valid_mask=None
mode='train', volume=data, lable=None, valid_mask=None
mode='test', volume=data, lable=data, valid_mask=data
mode='test', volume=data, lable=None, valid_mask=data
mode='test', volume=data, lable=data, valid_mask=None
mode='test', volume=data, lable=None, valid_mask=None
mode='val', volume=data, lable=data, valid_mask=data
mode='val', volume=data, lable=None, valid_mask=data
mode='val', volume=data, lable=data, valid_mask=None
mode='val', volume=data, lable=None, valid_mask=None