unreal.AnimGenAutoEncoder

class unreal.AnimGenAutoEncoder(outer: Object | None = None, name: Name | str = 'None')

Bases: DataAsset

AnimGen AutoEncoder

This asset can be used to learn a compact vector encoding of poses (and their velocities) in conjunction with the AnimDatabase. This compact vector encoding is useful for downstream machine learning tasks since it allows them to work in an abstract space rather than dealing directly with the pose representation.

C++ Source:

  • Plugin: AnimGen

  • Module: AnimGen

  • File: AnimGenAutoEncoder.h

Editor Properties: (see get_editor_property/set_editor_property)

  • decoder_inference_time (int32): [Read-Only] Current approximate inference time for the decoder in microseconds

  • decoder_size (int32): [Read-Only] Current size of the trained decoder in kilobytes

  • encoder_inference_time (int32): [Read-Only] Current approximate inference time for the encoder in microseconds

  • encoder_size (int32): [Read-Only] Current size of the trained encoder in kilobytes

  • trained_bone_locations (Array[Name]): [Read-Only] List of bone names where the translation is output by the auto-encoder

  • trained_bone_rotations (Array[Name]): [Read-Only] List of bone names where the rotation is output by the auto-encoder

  • trained_bone_scales (Array[Name]): [Read-Only] List of bone names where the scale is output by the auto-encoder

  • trained_frame_attributes (Array[AnimDatabaseFrameAttributeEntry]): [Read-Only] The Frame Attributes used for training

  • trained_skeleton (Skeleton): [Read-Only] The skeleton used for training

find_attribute_index(attribute_name) → int32

Finds the attribute index for a given name

Parameters:

attribute_name (Name)

Return type:

int32

find_bone_index(bone_name) → int32

Finds the AutoEncoder bone index corresponding to a bone name

Parameters:

bone_name (Name)

Return type:

int32

find_bone_indices(bone_in_bone_names_names) → Array[int32]

Finds the AutoEncoder bone indices corresponding to the given bone names

Parameters:

bone_in_bone_names_names (Array[Name])

Returns:

out_bone_indices (Array[int32]):

Return type:

Array[int32]

get_attribute_name(attribute_idx) → Name

Gets the name of the attribute with the given index

Parameters:

attribute_idx (int32)

Return type:

Name

get_attribute_names() → Array[Name]

Gets an array of all of the trained attribute names

Return type:

Array[Name]

get_attribute_num() → int32

Gets the number of attributes this AutoEncoder is trained on

Return type:

int32

get_attribute_type(attribute_idx) → AnimDatabaseAttributeType

Gets the type of the attribute with the given index

Parameters:

attribute_idx (int32)

Return type:

AnimDatabaseAttributeType

get_attribute_types() → Array[AnimDatabaseAttributeType]

Gets an array of all of the trained attribute types *

Return type:

Array[AnimDatabaseAttributeType]

get_bone_name(bone_idx) → Name

Gets the bone name for a given AutoEncoder bone index

Parameters:

bone_idx (int32)

Return type:

Name

get_bone_names() → Array[Name]

Gets the array of bone names for the AutoEncoder

Return type:

Array[Name]

get_bone_num() → int32

Gets the number of bones used by the AutoEncoder

Return type:

int32

get_bone_parent(bone_idx) → int32

Gets the parent for a given AutoEncoder bone index

Parameters:

bone_idx (int32)

Return type:

int32

get_bone_parents() → Array[int32]

Gets the array of bone parent indices for the AutoEncoder

Return type:

Array[int32]

get_content_hash() → int32

Gets a hash value which can be used to test if the underlying content of the auto-encoder networks has changed. This hash should not be relied on for detecting changes with certainty, and so should only be used for non-critical purposes (e.g. UI updates).

Return type:

int32

get_encoding_size() → int32

Gets the encoding size used by the AutoEncoder

Return type:

int32

get_pose_vector_size() → int32

Gets the pose vector size used by the AutoEncoder

Return type:

int32

invalidate() → None

Invalidate the AutoEncoder and reset back to a default state

is_valid() → bool

Check if the AutoEncoder is valid and trained

Return type:

bool