Update the gt segmentations of the data uploaded on Openneuro

[rohanbanerjee]

What is the issue?

The dataset has been uploaded to Openneuro and can be found here -- https://openneuro.org/datasets/ds005143/versions/1.2.0

• This issue is opened to discuss a strategy to update the segmentations of the subjects which have not been used for the training.
• The gt for the images whose ground truth has been used for training are precise and correct since they have been manually updated and QCd (ref issue Training and Inference discussion for baseline model #34 Training and inference discussion for active learning round 1 #35 Training and inference discussion for active learning round 2 #38 )
• For the rest of the subjects, there are cases of oversegmentation, undersegmentation and missing segmentations.

We should have a strategy to correct these gt.

Proposed strategy

Use our current latest model (unreleased as of now) to get the segmentations of the rest of the images.