poseutils.datasets.transformation package

Submodules

poseutils.datasets.transformation.CalculateMetrics module

class poseutils.datasets.transformation.CalculateMetrics.CalculateMetrics

Bases: poseutils.datasets.transformation.Transformation.Transformation

No-op Transformation class to indicate dataset metrics need to be recalculated.

poseutils.datasets.transformation.CropAndScale module

class poseutils.datasets.transformation.CropAndScale.CropAndScale(low=0, high=256, *args, **kwds)

Bases: poseutils.datasets.transformation.Transformation.Transformation

Class to apply crop and scale transformation. Makes call to poseutils.composite.scale_into_bounding_box_2d.

Parameters

  • low (int, optional) – Lowest value of the bounding box range, defaults to 0

  • high (int, optional) – Highest value of the bounding box range, defaults to 256

__call__(X, **kwds)

Applies transformation

Parameters

X (numpy.ndarray) – Joint positions (NxMx2), M = 14 or 16

Returns

Scaled joint positions (NxMx2), M = 14 or 16

Return type

numpy.ndarray

poseutils.datasets.transformation.Normalize module

class poseutils.datasets.transformation.Normalize.Normalize(skip_root=True)

Bases: poseutils.datasets.transformation.Transformation.Transformation

Applies z-score normalize transformation on the data.

Parameters

skip_root (bool, optional) – Whether to skip root/hip, defaults to True

__call__(X, mean, std, **kwds)

Applies transformation

\hat{\bm{X}} = \frac{\bm{X} - \bm{\mu}}{\bm{\sigma}}

Parameters

  • X (numpy.ndarray) – Joint positions (NxMxI), M = 14 or 16, I = 2 or 3

  • mean (numpy.ndarray) – Mean values to use when normalizing (MxI)

  • std (numpy.ndarray) – Standard deviation values to use when normalizing (MxI)

Returns

Transformed joint positions (NxMx2), M = 14 or 16

Return type

numpy.ndarray

poseutils.datasets.transformation.RootCenter module

class poseutils.datasets.transformation.RootCenter.RootCenter(root_idx=0)

Bases: poseutils.datasets.transformation.Transformation.Transformation

Subtracts the root/hip position from the rest of the joints

Parameters

root_idx (int, optional) – Root/hip index, defaults to 0

__call__(X, **kwds)

Applies tranfrormation

Parameters

X (numpy.ndarray) – Joint positions (NxMxI), M = 14 or 16, I = 2 or 3

Returns

Root/hip centered joint positions (NxMxI)

Return type

numpy.ndarray

poseutils.datasets.transformation.Transformation module

class poseutils.datasets.transformation.Transformation.Transformation

Bases: object

Base class for other Transformation classes.

poseutils.datasets.transformation.Unnormalize module

class poseutils.datasets.transformation.Unnormalize.Unnormalize(skip_root=True)

Bases: poseutils.datasets.transformation.Transformation.Transformation

Undoes z-score normalization transformation on the data.

Parameters

skip_root (bool, optional) – Whether to skip root/hip, defaults to True

__call__(X, mean, std, **kwds)

Applies transformation

\bm{X} = \hat{\bm{X}} \times \bm{\sigma} + \bm{\mu}

Parameters

  • X (numpy.ndarray) – Joint positions (NxMxI), M = 14 or 16, I = 2 or 3

  • mean (numpy.ndarray) – Mean values to use when normalizing (MxI)

  • std (numpy.ndarray) – Standard deviation values to use when normalizing (MxI)

Returns

Transformed joint positions (NxMx2), M = 14 or 16

Return type

numpy.ndarray

Module contents