pose_format.tensorflow.masked.tensor module

Classes:

MaskedTensor(tensor[, mask])

class pose_format.tensorflow.masked.tensor.MaskedTensor(tensor, mask=None)[source]

Bases: object

Methods:

arithmetic(action, other)

For element-wise arithmetic operations with another tensor.

div(other[, in_place, update_mask])

Divide tensor by another tensor.

fix_nan()

Replace NaN values with zeros while keeping mask.

float()

Convert tensor's data type to float32 while preserving mask.

gather(indexes)

Gather elements from tensor using indexes.

matmul(matrix)

Matrix multiplication a given matrix.

mean([axis])

Compute mean of tensor along a custom axis.

permute(dims)

Permute the dimensions of the tensor according to the provided tuple.

rename(*names)

Rename using custom names.

reshape(shape)

Reshape tensor into custom shape (tuple)

size(*args)

Get tensor's size along dimensions.

split(split_size_or_sections[, axis])

Split tensor

sqrt()

Element-wise square root of the tensor

square()

Element-wise square of the tensor.

squeeze(axis)

Remove dimensions with size 1 while updating the mask.

std([axis])

Compute the standard deviation of the tensor along the specified axis.

sum(axis)

Sum of tensor along specified axis while updating mask.

transpose(perm)

Transpose tensor according to given permutation.

variance([axis])

Compute variance of tensor along a specified axis

zero_filled()

Fill invalid values (as indicated by the mask) with zeros.

Parameters:
  • tensor (Tensor) –

  • mask (Tensor) –

arithmetic(action, other)[source]

For element-wise arithmetic operations with another tensor.

Parameters:
Returns:

A new MaskedTensor containing the result of the arithmetic operation.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

div(other, in_place=False, update_mask=True)[source]

Divide tensor by another tensor.

Parameters:
  • other (pose_format.tensorflow.masked.tensor.MaskedTensor) – The divisor tensor.

  • in_place (bool, optional) – Whether to do division in place. Default is False.

  • update_mask (bool, optional) – Whether to update mask after division. Default is True.

Returns:

Masked tensor after division.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

fix_nan()[source]

Replace NaN values with zeros while keeping mask.

Returns:

New MaskedTensor with NaN values replaced by zeros.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

float()[source]

Convert tensor’s data type to float32 while preserving mask.

Returns:

A new MaskedTensor with the tensor’s data type converted to float32.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

gather(indexes)[source]

Gather elements from tensor using indexes.

Parameters:

indexes (tf.Tensor or list or int) – Indexes used to select elements from tensor

Returns:

A new MaskedTensor containing elements gathered from the tensor using the indexes.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

matmul(matrix)[source]

Matrix multiplication a given matrix.

Parameters:

matrix (tf.Tensor) – Matrix to perform multiplication with.

Returns:

MaskedTensor` with result of matrix multiplication.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

mean(axis=None)[source]

Compute mean of tensor along a custom axis.

Parameters:

axis (None or int, optional) – Sxis along which to compute the mean. If None, compute the mean of the entire tensor. Default is None.

Returns:

The mean of the masked tensor.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

permute(dims)[source]

Permute the dimensions of the tensor according to the provided tuple.

Parameters:

dims (tuple) – The new order of dimensions after permutation.

Returns:

A new MaskedTensor with dimensions permuted according to the given tuple.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

rename(*names)[source]

Rename using custom names.

Parameters:

*names (str) – New names of the dimensions.

Returns:

A new MaskedTensor with dimensions renamed.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

reshape(shape)[source]

Reshape tensor into custom shape (tuple)

Parameters:

shape (tuple) – New shape of tensor.

Returns:

new MaskedTensor with specified shape.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

size(*args)[source]

Get tensor’s size along dimensions.

Parameters:

*args (int) – Dimensions for which to get size

Returns:

Size of tensor of specified dimensions.

Return type:

int or tuple of int

split(split_size_or_sections, axis=0)[source]

Split tensor

Parameters:
  • split_size_or_sections (int or tf.Tensor) – Number of splits or sizes of each split/sections.

  • axis (int, optional) – Axis along which to do the splitting. Default is 0.

Returns:

List of new MaskedTensor objects containing the splits.

Return type:

list of pose_format.tensorflow.masked.tensor.MaskedTensor

sqrt()[source]

Element-wise square root of the tensor

Returns:

A new MaskedTensor containing the square root values of the original tensor.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

square()[source]

Element-wise square of the tensor.

Returns:

A new MaskedTensor containing the squared values of the original tensor.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

squeeze(axis)[source]

Remove dimensions with size 1 while updating the mask.

Parameters:

axis (int or None) – The axis along which to perform squeezing.

Returns:

MaskedTensor` with dimensions removed and mask updated.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

std(axis=None)[source]

Compute the standard deviation of the tensor along the specified axis.

Parameters:

axis (None or int, optional) – The axis along which to compute the standard deviation. If None, compute the standard deviation of the entire tensor. Default is None.

Returns:

The standard deviation of the tensor.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

sum(axis)[source]

Sum of tensor along specified axis while updating mask.

Parameters:

axis (int or None) – Axis along which to compute sum. If None, compute the sum over all elements.

Returns:

A new MaskedTensor containing the sums of the tensor along the specified axis.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

transpose(perm)[source]

Transpose tensor according to given permutation.

Parameters:

perm (List[int]) – The new order of dimensions/permutation after transposition.

Returns:

MaskedTensor with dimensions transposed according to the given permutation.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

variance(axis=None)[source]

Compute variance of tensor along a specified axis

Parameters:

axis (None or int, optional) – Axis along which to compute the variance. If None, compute the variance of the entire tensor. Default is None.

Returns:

The variance of the masked tensor.

Return type:

pose_format.tensorflow.masked.tensor.MaskedTensor

zero_filled()[source]

Fill invalid values (as indicated by the mask) with zeros.

Returns:

Tensor with the same shape as self.tensor but with zeros where the mask is False.

Return type:

tf.Tensor