pose_format.torch.representation.points module

Classes:

PointsRepresentation(*args, **kwargs)

Class to represent points in a tensor format for processing.

class pose_format.torch.representation.points.PointsRepresentation(*args, **kwargs)

Bases: Module

Class to represent points in a tensor format for processing.

Methods:

forward(p1s)

Transforms input tensor representing points into a desired tensor format.

Attributes:

training

forward(p1s)

Transforms input tensor representing points into a desired tensor format.

The transformation process with zero-filling the masked values in input tensor and reshaping tensor by transposing its dimensions to match the desired output format.

Parameters:

p1s (MaskedTensor) – Tensor representing a set of points. Shape: (Points, Batch, Len, Dims).

Returns:

Transformed tensor representing the points. Shape: (Points*Dims, Batch, Len).

Return type:

torch.Tensor

Note

This method first fills masked values in input tensor with zeros. Then, it reshapes tensor by transposing dimensions to match its desired output format

training: bool