Feature Extraction

class brainlit.feature_extraction.NeighborhoodFeatures(url: str, radius: List[int] = [1, 1, 1], offset: List[int] = [15, 15, 15], segment_url: Optional[str] = None)

Computes features based off neighborhood properties.

Parameters
  • url -- Precompued path either to a file URI or url URI of image data.

  • radius -- The radius around each point considered a neighborhood, in each dimension. If radius is [x,y,z], the neighborhood will be a [2x+1, 2y+1, 2z+1] volume centered at the point of interest. Defaults to [1, 1, 1].

  • offset -- Added to the coordinates of a positive sample to generate a negative sample. Defaults to [15, 15, 15].

  • segment_url -- Precompued path either to a file URI or url URI of segmentation data.

url

CloudVolumePrecomputedPath to image data.

size

A size hyperparameter. In Neighborhoods, this is the radius.

offset

Added to the coordinates of a positive sample to generate a negative sample.

download_time

Tracks time taken to download the data.

conversion_time

Tracks time taken to convert data to features.

write_time

Tracks time taken to write features to files.

segment_url

CloudVolumePrecomputedPath to segmentation data.