vertebrae.segmentation

Spatial feature alignment and token materialization for segmentation.

Classes

SegmentationMaterialization

One flattened, grouped segmentation embedding output.

Functions

`materialize_segmentation_outputs`(dataset, extractor[, ...])	Extract, align, sample, and flatten all declared spatial outputs.
`iter_materialize_segmentation_outputs`(dataset, extractor)	Yield flattened spatial outputs one at a time after shared extraction.

Module Contents

class vertebrae.segmentation.SegmentationMaterialization[source]: One flattened, grouped segmentation embedding output.

vertebrae.segmentation.materialize_segmentation_outputs(dataset, extractor, config=None, batch_size=16, resource_profiler=None, memory_config=None)[source]

Extract, align, sample, and flatten all declared spatial outputs.

Parameters:

dataset (vertebrae.datasets.segmentation.SegmentationDataset)
extractor (Any)
config (Optional[vertebrae.config.SegmentationConfig])
batch_size (int)
resource_profiler (Optional[Any])
memory_config (Optional[vertebrae.config.MemoryConfig])

Return type:

List[SegmentationMaterialization]

vertebrae.segmentation.iter_materialize_segmentation_outputs(dataset, extractor, config=None, batch_size=16, resource_profiler=None, memory_config=None)[source]

Yield flattened spatial outputs one at a time after shared extraction.

Parameters:

dataset (vertebrae.datasets.segmentation.SegmentationDataset)
extractor (Any)
config (Optional[vertebrae.config.SegmentationConfig])
batch_size (int)
resource_profiler (Optional[Any])
memory_config (Optional[vertebrae.config.MemoryConfig])

Return type:

Iterator[SegmentationMaterialization]