pose_format.torch package

PyTorch-specific functionalities.

Subpackages:

• pose_format.torch.masked package
  • pose_format.torch.masked.tensor module
    • MaskedTensor
      • MaskedTensor.arithmetic()
      • MaskedTensor.cuda()
      • MaskedTensor.div()
      • MaskedTensor.fix_nan()
      • MaskedTensor.matmul()
      • MaskedTensor.permute()
      • MaskedTensor.pow_()
      • MaskedTensor.rename()
      • MaskedTensor.reshape()
      • MaskedTensor.size()
      • MaskedTensor.split()
      • MaskedTensor.squeeze()
      • MaskedTensor.sum()
      • MaskedTensor.to()
      • MaskedTensor.transpose()
      • MaskedTensor.zero_filled()
  • pose_format.torch.masked.torch module
    • MaskedTorch
      • MaskedTorch.cat()
      • MaskedTorch.squeeze()
      • MaskedTorch.stack()
      • MaskedTorch.zeros()
    • TorchFallback
      • TorchFallback.doesnt_change_mask
  • pose_format.torch.masked.torch_test module
    • TestMaskedTorch
      • TestMaskedTorch.test_cat()
      • TestMaskedTorch.test_not_implemented_method()
      • TestMaskedTorch.test_stack()
      • TestMaskedTorch.test_zeros_mask_value()
      • TestMaskedTorch.test_zeros_tensor_shape()
      • TestMaskedTorch.test_zeros_tensor_type_bool()
      • TestMaskedTorch.test_zeros_tensor_type_float()
      • TestMaskedTorch.test_zeros_tensor_value()
  • pose_format.torch.masked.tensor module
    • MaskedTensor
      • MaskedTensor.arithmetic()
      • MaskedTensor.cuda()
      • MaskedTensor.div()
      • MaskedTensor.fix_nan()
      • MaskedTensor.matmul()
      • MaskedTensor.permute()
      • MaskedTensor.pow_()
      • MaskedTensor.rename()
      • MaskedTensor.reshape()
      • MaskedTensor.size()
      • MaskedTensor.split()
      • MaskedTensor.squeeze()
      • MaskedTensor.sum()
      • MaskedTensor.to()
      • MaskedTensor.transpose()
      • MaskedTensor.zero_filled()
  • pose_format.torch.masked.torch module
    • MaskedTorch
      • MaskedTorch.cat()
      • MaskedTorch.squeeze()
      • MaskedTorch.stack()
      • MaskedTorch.zeros()
    • TorchFallback
      • TorchFallback.doesnt_change_mask
  • pose_format.torch.masked.torch_test module
    • TestMaskedTorch
      • TestMaskedTorch.test_cat()
      • TestMaskedTorch.test_not_implemented_method()
      • TestMaskedTorch.test_stack()
      • TestMaskedTorch.test_zeros_mask_value()
      • TestMaskedTorch.test_zeros_tensor_shape()
      • TestMaskedTorch.test_zeros_tensor_type_bool()
      • TestMaskedTorch.test_zeros_tensor_type_float()
      • TestMaskedTorch.test_zeros_tensor_value()
• pose_format.torch.representation package
  • pose_format.torch.representation.angle module
    • AngleRepresentation
      • AngleRepresentation.forward()
      • AngleRepresentation.training
  • pose_format.torch.representation.angle_test module
    • TestAngleRepresentation
      • TestAngleRepresentation.test_call_masked_value_should_be_zero()
      • TestAngleRepresentation.test_call_value_should_be_angle()
  • pose_format.torch.representation.distance module
    • DistanceRepresentation
      • DistanceRepresentation.distance()
      • DistanceRepresentation.forward()
      • DistanceRepresentation.training
  • pose_format.torch.representation.distance_test module
    • TestDistanceRepresentation
      • TestDistanceRepresentation.test_call_masked_value_should_be_zero()
      • TestDistanceRepresentation.test_call_value_should_be_distance()
  • pose_format.torch.representation.inner_angle module
    • InnerAngleRepresentation
      • InnerAngleRepresentation.forward()
      • InnerAngleRepresentation.training
    • get_vectors_norm()
  • pose_format.torch.representation.inner_angle_test module
    • TestInnerAngleRepresentation
      • TestInnerAngleRepresentation.test_call_masked_value_should_be_zero()
      • TestInnerAngleRepresentation.test_call_value_should_be_inner_angle()
  • pose_format.torch.representation.point_line_distance module
    • PointLineDistanceRepresentation
      • PointLineDistanceRepresentation.forward()
      • PointLineDistanceRepresentation.training
  • pose_format.torch.representation.point_line_distance_test module
    • TestPointLineDistanceRepresentation
      • TestPointLineDistanceRepresentation.test_call_masked_value_should_be_zero()
      • TestPointLineDistanceRepresentation.test_call_value_should_be_distance()
  • pose_format.torch.representation.points module
    • PointsRepresentation
      • PointsRepresentation.forward()
      • PointsRepresentation.training

Submodules

pose_format.torch.pose_body module

Classes:

TorchPoseBody(fps, data, confidence)

TorchPoseBody class of pose information with PyTorch tensors.

class pose_format.torch.pose_body.TorchPoseBody(fps, data, confidence)

Bases: PoseBody

TorchPoseBody class of pose information with PyTorch tensors.

This class extends the PoseBody class and provides methods for manipulating pose data using PyTorch tensors.

Methods:

cuda()	Move data and cofidence of tensors to GPU
flatten()	Flatten pose data along the associated confidence values.
get_points(indexes)	Get specific points from pose data.
matmul(matrix)	Matrix multiplication on pose data.
points_perspective()	Get pose data with dimensions permuted according to POINTS_DIMS.
zero_filled()	Fill invalid values with zeros.

Attributes:

tensor_reader

Parameters:

• fps (float) –

• data (MaskedTensor | Tensor) –

• confidence (Tensor) –

cuda()

Move data and cofidence of tensors to GPU

flatten()

Flatten pose data along the associated confidence values.

Returns:

Flattened tensor containing indexes, confidence values, and data.

Return type:

torch.Tensor

get_points(indexes)

Get specific points from pose data.

Parameters:

indexes (List[int]) – List of indexes specifying the points that you need.

Returns:

New TorchPoseBody instance containing specified points and associated confidence values.

Return type:

TorchPoseBody

matmul(matrix)

Matrix multiplication on pose data.

Parameters:

matrix (np.ndarray) – matrix to perform multiplication with

Returns:

A new TorchPoseBody instance with results of matrix multiplication.

Return type:

TorchPoseBody

points_perspective()

Get pose data with dimensions permuted according to POINTS_DIMS.

Returns:

A MaskedTensor instance with dimensions permuted for points perspective.

Return type:

MaskedTensor

tensor_reader = 'unpack_torchstr: Reader format for unpacking Torch tensors.'

zero_filled()

Fill invalid values with zeros.

Returns:

TorchPoseBody instance with masked data filled with zeros.

Return type:

TorchPoseBody

pose_format.torch.pose_representation module

Classes:

TorchPoseRepresentation(header[, ...])

TorchPoseRepresentation class representing pose information using PyTorch tensors.

class pose_format.torch.pose_representation.TorchPoseRepresentation(header, rep_modules1=[], rep_modules2=[], rep_modules3=[])

Bases: PoseRepresentation

TorchPoseRepresentation class representing pose information using PyTorch tensors.

This class extends the PoseRepresentation class and provides methods for manipulating and representing pose data using PyTorch tensors.

Parameters:

• header (PoseHeader) – Header describing the pose data structure.

• rep_modules1 (List) – List of additional representation modules (level 1) to apply to pose data.

• rep_modules2 (List) – List of additional representation modules (level 2) to apply to pose data.

• rep_modules3 (List) – List of additional representation modules (level 3) to apply to pose data.

Methods:

group_embeds(embeds)	Group and reshape embedded tensors for batch processing.
permute(src, shape)	Permute dimensions of tensor according to a specified shape (tuple).

group_embeds(embeds)

Group and reshape embedded tensors for batch processing.

Parameters:

embeds (List[torch.Tensor]) – List of embedded tensors of size (embed_size, Batch, Len).

Returns:

A tensor of size (Batch, Len, embed_size) with grouped and reshaped embedded tensors.

Return type:

torch.Tensor

permute(src, shape)

Permute dimensions of tensor according to a specified shape (tuple).

Parameters:

• src (torch.Tensor) – tensor to permute

• shape (tuple) – desired shape of the tensor after permutation.

Returns:

tensor with permuted dimensions according to specified shape.

Return type:

torch.Tensor