Fixes #365.

Initial proposition for inverting deterministic spatial transforms.

Status

Work in progress

Types of changes

• Non-breaking change (fix or new feature that would not break existing functionality).
• Breaking change (fix or new feature that would cause existing functionality to change).
• New tests added to cover the changes.
• Integration tests passed locally by running ./runtests.sh --codeformat --coverage.
• Quick tests passed locally by running ./runtests.sh --quick.
• In-line docstrings updated.
• Documentation updated, tested make html command in the docs/ folder.

To do

• Implement inverse method for a deterministic spatial transform (SpatialPadd)
• Implement inverse for Compose class
• Create serialised representation of transform (append_applied_transforms and get_input_args)
• Implement for a random class (to check we can safely store random parameters)
• Test with CacheDataset and DataLoader to check for multiprocessing problems
• Check GPU vs CPU
• Check numpy vs. tensor
• Check inclusion of batch dimension
• Test with transforms that return lists (e.g., RandSpatialCropSamplesd)
• Add to transforms.rst

Invertible classes

• Compose
• SpatialPadd
• BorderPadd
• DivisiblePadd
• SpatialCropd
• CenterSpatialCropd
• CropForegroundd
• ResizeWithPadOrCropd
• Spacingd
• Orientationd
• Rotate90d
• Resized
• Flipd
• Rotated
• Zoomd
• RandSpatialCropd
• RandSpatialCropSamplesd
• RandWeightedCropd
• RandCropByPosNegLabeld
• RandRotate90d
• RandAffined
• Rand2DElasticd
• Rand3DElasticd
• RandFlipd
• RandRotated
• RandZoomd

Not applicable

• BoundingRectd -- doesn't modify the image, but adds a dictionary entry containing the bounding box.