Datatypes

`rerun.datatypes`

`class Angle`

Bases: AngleExt

Datatype: Angle in either radians or degrees.

`inner: float = field(converter=float)` `class-attribute` `instance-attribute`

Radians (float): 3D rotation angle in radians. Only one of degrees or radians should be set.

Degrees (float): 3D rotation angle in degrees. Only one of degrees or radians should be set.

`def init(rad=None, deg=None)`

Create a new instance of the Angle datatype.

PARAMETER	DESCRIPTION
`rad`	Angle in radians, specify either `rad` or `deg`. TYPE: `float \| None` DEFAULT: `None`
`deg`	Angle in degrees, specify either `rad` or `deg`. TYPE: `float \| None` DEFAULT: `None`

`class AnnotationInfo`

Bases: AnnotationInfoExt

Datatype: Annotation info annotating a class id or key-point id.

Color and label will be used to annotate entities/keypoints which reference the id. The id refers either to a class or key-point id

`def init(id, label=None, color=None)`

Create a new instance of the AnnotationInfo datatype.

PARAMETER	DESCRIPTION
`id`	`ClassId` or `KeypointId` to which this annotation info belongs. TYPE: `int`
`label`	The label that will be shown in the UI. TYPE: `Utf8Like \| None` DEFAULT: `None`
`color`	The color that will be applied to the annotated entity. TYPE: `Rgba32Like \| None` DEFAULT: `None`

`class ClassDescription`

Bases: ClassDescriptionExt

Datatype: The description of a semantic Class.

If an entity is annotated with a corresponding ClassId, rerun will use the attached AnnotationInfo to derive labels and colors.

Keypoints within an annotation class can similarly be annotated with a KeypointId in which case we should defer to the label and color for the AnnotationInfo specifically associated with the Keypoint.

Keypoints within the class can also be decorated with skeletal edges. Keypoint-connections are pairs of KeypointIds. If an edge is defined, and both keypoints exist within the instance of the class, then the keypoints should be connected with an edge. The edge should be labeled and colored as described by the class's AnnotationInfo.

Note that a ClassDescription can be directly logged using rerun.log. This is equivalent to logging a rerun.AnnotationContext containing a single ClassDescription.

`def init(*, info, keypoint_annotations=[], keypoint_connections=[])`

Create a new instance of the ClassDescription datatype.

PARAMETER	DESCRIPTION
`info`	The `AnnotationInfo` for the class. TYPE: `AnnotationInfoLike`
`keypoint_annotations`	The `AnnotationInfo` for all the keypoints. TYPE: `Sequence[AnnotationInfoLike] \| None` DEFAULT: `[]`
`keypoint_connections`	The connections between keypoints. TYPE: `Sequence[KeypointPairLike] \| None` DEFAULT: `[]`

`class ClassDescriptionMapElem`

Bases: ClassDescriptionMapElemExt

Datatype: A helper type for mapping class IDs to class descriptions.

This is internal to the AnnotationContext structure.

`def init(class_id, class_description)`

Create a new instance of the ClassDescriptionMapElem datatype.

PARAMETER	DESCRIPTION
`class_id`	The key: the class ID. TYPE: `ClassIdLike`
`class_description`	The value: class name, color, etc. TYPE: `ClassDescriptionLike`

`class ClassId`

Bases: ClassIdExt

Datatype: A 16-bit ID representing a type of semantic class.

`def init(id)`

Create a new instance of the ClassId datatype.

`class Float32`

Datatype: A single-precision 32-bit IEEE 754 floating point number.

`def init(value)`

Create a new instance of the Float32 datatype.

`class KeypointId`

Bases: KeypointIdExt

Datatype: A 16-bit ID representing a type of semantic keypoint within a class.

KeypointIds are only meaningful within the context of a [rerun.datatypes.ClassDescription].

Used to look up an [rerun.datatypes.AnnotationInfo] for a Keypoint within the [rerun.components.AnnotationContext].

`def init(id)`

Create a new instance of the KeypointId datatype.

`class KeypointPair`

Bases: KeypointPairExt

Datatype: A connection between two Keypoints.

`def init(keypoint0, keypoint1)`

Create a new instance of the KeypointPair datatype.

PARAMETER	DESCRIPTION
`keypoint0`	The first point of the pair. TYPE: `KeypointIdLike`
`keypoint1`	The second point of the pair. TYPE: `KeypointIdLike`

`class Mat3x3`

Bases: Mat3x3Ext

Datatype: A 3x3 Matrix.

Matrices in Rerun are stored as flat list of coefficients in column-major order:

            column 0       column 1       column 2
       -------------------------------------------------
row 0 | flat_columns[0] flat_columns[3] flat_columns[6]
row 1 | flat_columns[1] flat_columns[4] flat_columns[7]
row 2 | flat_columns[2] flat_columns[5] flat_columns[8]

However, construction is done from a list of rows, which follows NumPy's convention:

np.testing.assert_array_equal(
    rr.datatypes.Mat3x3([1, 2, 3, 4, 5, 6, 7, 8, 9]).flat_columns, np.array([1, 4, 7, 2, 5, 8, 3, 6, 9], dtype=np.float32)
)
np.testing.assert_array_equal(
    rr.datatypes.Mat3x3([[1, 2, 3], [4, 5, 6], [7, 8, 9]]).flat_columns,
    np.array([1, 4, 7, 2, 5, 8, 3, 6, 9], dtype=np.float32),
)

If you want to construct a matrix from a list of columns instead, use the named columns parameter:

np.testing.assert_array_equal(
    rr.datatypes.Mat3x3(columns=[1, 2, 3, 4, 5, 6, 7, 8, 9]).flat_columns,
    np.array([1, 2, 3, 4, 5, 6, 7, 8, 9], dtype=np.float32),
)
np.testing.assert_array_equal(
    rr.datatypes.Mat3x3(columns=[[1, 2, 3], [4, 5, 6], [7, 8, 9]]).flat_columns,
    np.array([1, 2, 3, 4, 5, 6, 7, 8, 9], dtype=np.float32),
)

`def init(rows=None, *, columns=None)`

`class Mat4x4`

Bases: Mat4x4Ext

Datatype: A 4x4 Matrix.

Matrices in Rerun are stored as flat list of coefficients in column-major order:

           column 0         column 1         column 2         column 3
       --------------------------------------------------------------------
row 0 | flat_columns[0]  flat_columns[4]  flat_columns[8]  flat_columns[12]
row 1 | flat_columns[1]  flat_columns[5]  flat_columns[9]  flat_columns[13]
row 2 | flat_columns[2]  flat_columns[6]  flat_columns[10] flat_columns[14]
row 3 | flat_columns[3]  flat_columns[7]  flat_columns[11] flat_columns[15]

However, construction is done from a list of rows, which follows NumPy's convention:

np.testing.assert_array_equal(
    rr.datatypes.Mat4x4([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16]).flat_columns,
    np.array([1, 5, 9, 13, 2, 6, 10, 14, 3, 7, 11, 15, 4, 8, 12, 16], dtype=np.float32),
)
np.testing.assert_array_equal(
    rr.datatypes.Mat4x4([[1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], [13, 14, 15, 16]]).flat_columns,
    np.array([1, 5, 9, 13, 2, 6, 10, 14, 3, 7, 11, 15, 4, 8, 12, 16], dtype=np.float32),
)

If you want to construct a matrix from a list of columns instead, use the named columns parameter:

np.testing.assert_array_equal(
    rr.datatypes.Mat4x4(columns=[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16]).flat_columns,
    np.array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16], dtype=np.float32),
)
np.testing.assert_array_equal(
    rr.datatypes.Mat4x4(columns=[[1, 2, 3, 4], [5, 6, 7, 8], [9, 10, 11, 12], [13, 14, 15, 16]]).flat_columns,
    np.array([1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16], dtype=np.float32),
)

`def init(rows=None, *, columns=None)`

`class Material`

Bases: MaterialExt

Datatype: Material properties of a mesh.

`def init(albedo_factor=None)`

Create a new instance of the Material datatype.

PARAMETER	DESCRIPTION
`albedo_factor`	Optional color multiplier. TYPE: `Rgba32Like \| None` DEFAULT: `None`

`class MeshProperties`

Bases: MeshPropertiesExt

Datatype: Optional triangle indices for a mesh.

`def init(indices=None)`

Create a new instance of the MeshProperties datatype.

PARAMETER	DESCRIPTION
`indices`	A flattened array of vertex indices that describe the mesh's triangles. Its length must be divisible by 3. TYPE: `ArrayLike \| None` DEFAULT: `None`

`class Quaternion`

Bases: QuaternionExt

Datatype: A Quaternion represented by 4 real numbers.

Note: although the x,y,z,w components of the quaternion will be passed through to the datastore as provided, when used in the viewer Quaternions will always be normalized.

`class Rgba32`

Bases: Rgba32Ext

Datatype: An RGBA color with unmultiplied/separate alpha, in sRGB gamma space with linear alpha.

The color is stored as a 32-bit integer, where the most significant byte is R and the least significant byte is A.

Float colors are assumed to be in 0-1 gamma sRGB space. All other colors are assumed to be in 0-255 gamma sRGB space. If there is an alpha, we assume it is in linear space, and separate (NOT pre-multiplied).

`def init(rgba)`

Create a new instance of the Rgba32 datatype.

`class Rotation3D`

Bases: Rotation3DExt

Datatype: A 3D rotation.

`inner: datatypes.Quaternion | datatypes.RotationAxisAngle = field(converter=Rotation3DExt.inner__field_converter_override)` `class-attribute` `instance-attribute`

Quaternion (datatypes.Quaternion): Rotation defined by a quaternion.

AxisAngle (datatypes.RotationAxisAngle): Rotation defined with an axis and an angle.

`class RotationAxisAngle`

Bases: RotationAxisAngleExt

Datatype: 3D rotation represented by a rotation around a given axis.

`def init(axis, angle=None, *, radians=None, degrees=None)`

Create a new instance of the RotationAxisAngle datatype.

PARAMETER	DESCRIPTION
`axis`	Axis to rotate around. This is not required to be normalized. If normalization fails (typically because the vector is length zero), the rotation is silently ignored. TYPE: `Vec3DLike`
`angle`	How much to rotate around the axis. TYPE: `AngleLike \| None` DEFAULT: `None`
`radians`	How much to rotate around the axis, in radians. Specify this instead of `degrees` or `angle`. TYPE: `float \| None` DEFAULT: `None`
`degrees`	How much to rotate around the axis, in radians. Specify this instead of `radians` or `angle`. TYPE: `float \| None` DEFAULT: `None`

`class Scale3D`

Bases: Scale3DExt

Datatype: 3D scaling factor, part of a transform representation.

Example

# uniform scaling
scale = rr.datatypes.Scale3D(3.)

# non-uniform scaling
scale = rr.datatypes.Scale3D([1, 1, -1])
scale = rr.datatypes.Scale3D(rr.datatypes.Vec3D([1, 1, -1]))

`inner: datatypes.Vec3D | float = field(converter=Scale3DExt.inner__field_converter_override)` `class-attribute` `instance-attribute`

ThreeD (datatypes.Vec3D): Individual scaling factors for each axis, distorting the original object.

Uniform (float): Uniform scaling factor along all axis.

`class TensorBuffer`

Bases: TensorBufferExt

Datatype: The underlying storage for a Tensor.

Tensor elements are stored in a contiguous buffer of a single type.

`inner: npt.NDArray[np.float16] | npt.NDArray[np.float32] | npt.NDArray[np.float64] | npt.NDArray[np.int16] | npt.NDArray[np.int32] | npt.NDArray[np.int64] | npt.NDArray[np.int8] | npt.NDArray[np.uint16] | npt.NDArray[np.uint32] | npt.NDArray[np.uint64] | npt.NDArray[np.uint8] = field(converter=TensorBufferExt.inner__field_converter_override)` `class-attribute` `instance-attribute`

U8 (npt.NDArray[np.uint8]):

U16 (npt.NDArray[np.uint16]):

U32 (npt.NDArray[np.uint32]):

U64 (npt.NDArray[np.uint64]):

I8 (npt.NDArray[np.int8]):

I16 (npt.NDArray[np.int16]):

I32 (npt.NDArray[np.int32]):

I64 (npt.NDArray[np.int64]):

F16 (npt.NDArray[np.float16]):

F32 (npt.NDArray[np.float32]):

F64 (npt.NDArray[np.float64]):

JPEG (npt.NDArray[np.uint8]):

`class TensorData`

Bases: TensorDataExt

Datatype: A multi-dimensional Tensor of data.

The number of dimensions and their respective lengths is specified by the shape field. The dimensions are ordered from outermost to innermost. For example, in the common case of a 2D RGB Image, the shape would be [height, width, channel].

These dimensions are combined with an index to look up values from the buffer field, which stores a contiguous array of typed values.

`def init(*, shape=None, buffer=None, array=None, dim_names=None)`

Construct a TensorData object.

The TensorData object is internally represented by three fields: shape and buffer.

This constructor provides additional arguments 'array', and 'dim_names'. When passing in a multi-dimensional array such as a np.ndarray, the shape and buffer fields will be populated automagically.

PARAMETER	DESCRIPTION
`self`	The TensorData object to construct. TYPE: `Any`
`shape`	The shape of the tensor. If None, and an array is proviced, the shape will be inferred from the shape of the array. TYPE: `Sequence[TensorDimensionLike] \| None` DEFAULT: `None`
`buffer`	The buffer of the tensor. If None, and an array is provided, the buffer will be generated from the array. TYPE: `TensorBufferLike \| None` DEFAULT: `None`
`array`	A numpy array (or The array of the tensor. If None, the array will be inferred from the buffer. TYPE: `TensorLike \| None` DEFAULT: `None`
`dim_names`	The names of the tensor dimensions when generating the shape from an array. TYPE: `Sequence[str \| None] \| None` DEFAULT: `None`

`class TensorDimension`

Datatype: A single dimension within a multi-dimensional tensor.

`def init(size, name=None)`

Create a new instance of the TensorDimension datatype.

PARAMETER	DESCRIPTION
`size`	The length of this dimension. TYPE: `int`
`name`	The name of this dimension, e.g. "width", "height", "channel", "batch', …. TYPE: `str \| None` DEFAULT: `None`

`class Transform3D`

Bases: Transform3DExt

Datatype: Representation of a 3D affine transform.

`inner: datatypes.TranslationAndMat3x3 | datatypes.TranslationRotationScale3D = field()` `class-attribute` `instance-attribute`

TranslationAndMat3x3 (datatypes.TranslationAndMat3x3):

TranslationRotationScale (datatypes.TranslationRotationScale3D):

`class TranslationAndMat3x3`

Bases: TranslationAndMat3x3Ext

Datatype: Representation of an affine transform via a 3x3 affine matrix paired with a translation.

First applies the matrix, then the translation.

`def init(translation=None, mat3x3=None, *, from_parent=False)`

Create a new instance of the TranslationAndMat3x3 datatype.

PARAMETER	DESCRIPTION
`translation`	3D translation, applied after the matrix. TYPE: `Vec3DLike \| None` DEFAULT: `None`
`mat3x3`	3x3 matrix for scale, rotation & shear. TYPE: `Mat3x3Like \| None` DEFAULT: `None`
`from_parent`	If true, the transform maps from the parent space to the space where the transform was logged. Otherwise, the transform maps from the space to its parent. TYPE: `bool` DEFAULT: `False`

`class TranslationRotationScale3D`

Bases: TranslationRotationScale3DExt

Datatype: Representation of an affine transform via separate translation, rotation & scale.

`class UVec2D`

Datatype: A uint32 vector in 2D space.

`def init(xy)`

Create a new instance of the UVec2D datatype.

`class UVec3D`

Datatype: A uint32 vector in 3D space.

`def init(xyz)`

Create a new instance of the UVec3D datatype.

`class UVec4D`

Datatype: A uint vector in 4D space.

`def init(xyzw)`

Create a new instance of the UVec4D datatype.

`class Utf8`

Bases: Utf8Ext

Datatype: A string of text, encoded as UTF-8.

`def init(value)`

Create a new instance of the Utf8 datatype.

`class Vec2D`

Bases: Vec2DExt

Datatype: A vector in 2D space.

`def init(xy)`

Create a new instance of the Vec2D datatype.

`class Vec3D`

Bases: Vec3DExt

Datatype: A vector in 3D space.

`def init(xyz)`

Create a new instance of the Vec3D datatype.

`class Vec4D`

Bases: Vec4DExt

Datatype: A vector in 4D space.

`def init(xyzw)`

Create a new instance of the Vec4D datatype.

Datatypes

rerun.datatypes

class Angle

inner: float = field(converter=float) class-attribute instance-attribute

def __init__(rad=None, deg=None)

class AnnotationInfo

def __init__(id, label=None, color=None)

class ClassDescription

def __init__(*, info, keypoint_annotations=[], keypoint_connections=[])

class ClassDescriptionMapElem

def __init__(class_id, class_description)

class ClassId

def __init__(id)

class Float32

def __init__(value)

class KeypointId

def __init__(id)

class KeypointPair

def __init__(keypoint0, keypoint1)

class Mat3x3

def __init__(rows=None, *, columns=None)

class Mat4x4

def __init__(rows=None, *, columns=None)

class Material

def __init__(albedo_factor=None)

class MeshProperties

def __init__(indices=None)

class Quaternion

class Rgba32

def __init__(rgba)

class Rotation3D

inner: datatypes.Quaternion | datatypes.RotationAxisAngle = field(converter=Rotation3DExt.inner__field_converter_override) class-attribute instance-attribute

class RotationAxisAngle

def __init__(axis, angle=None, *, radians=None, degrees=None)

class Scale3D

Example

inner: datatypes.Vec3D | float = field(converter=Scale3DExt.inner__field_converter_override) class-attribute instance-attribute

class TensorBuffer

class TensorData

def __init__(*, shape=None, buffer=None, array=None, dim_names=None)

class TensorDimension

def __init__(size, name=None)

class Transform3D

inner: datatypes.TranslationAndMat3x3 | datatypes.TranslationRotationScale3D = field() class-attribute instance-attribute

class TranslationAndMat3x3

def __init__(translation=None, mat3x3=None, *, from_parent=False)

class TranslationRotationScale3D

class UVec2D

def __init__(xy)

class UVec3D

def __init__(xyz)

class UVec4D

def __init__(xyzw)

class Utf8

def __init__(value)

class Vec2D

def __init__(xy)

class Vec3D

def __init__(xyz)

class Vec4D

def __init__(xyzw)

`rerun.datatypes`

`class Angle`

`inner: float = field(converter=float)` `class-attribute` `instance-attribute`

`def init(rad=None, deg=None)`

`class AnnotationInfo`

`def init(id, label=None, color=None)`

`class ClassDescription`

`def init(*, info, keypoint_annotations=[], keypoint_connections=[])`

`class ClassDescriptionMapElem`

`def init(class_id, class_description)`

`class ClassId`

`def init(id)`

`class Float32`

`def init(value)`

`class KeypointId`

`def init(id)`

`class KeypointPair`

`def init(keypoint0, keypoint1)`

`class Mat3x3`

`def init(rows=None, *, columns=None)`

`class Mat4x4`

`def init(rows=None, *, columns=None)`

`class Material`

`def init(albedo_factor=None)`

`class MeshProperties`

`def init(indices=None)`

`class Quaternion`

`class Rgba32`

`def init(rgba)`

`class Rotation3D`

`inner: datatypes.Quaternion | datatypes.RotationAxisAngle = field(converter=Rotation3DExt.inner__field_converter_override)` `class-attribute` `instance-attribute`

`class RotationAxisAngle`

`def init(axis, angle=None, *, radians=None, degrees=None)`

`class Scale3D`

`inner: datatypes.Vec3D | float = field(converter=Scale3DExt.inner__field_converter_override)` `class-attribute` `instance-attribute`

`class TensorBuffer`

`class TensorData`

`def init(*, shape=None, buffer=None, array=None, dim_names=None)`

`class TensorDimension`

`def init(size, name=None)`

`class Transform3D`

`inner: datatypes.TranslationAndMat3x3 | datatypes.TranslationRotationScale3D = field()` `class-attribute` `instance-attribute`

`class TranslationAndMat3x3`

`def init(translation=None, mat3x3=None, *, from_parent=False)`

`class TranslationRotationScale3D`

`class UVec2D`

`def init(xy)`

`class UVec3D`

`def init(xyz)`

`class UVec4D`

`def init(xyzw)`

`class Utf8`

`def init(value)`

`class Vec2D`

`def init(xy)`

`class Vec3D`

`def init(xyz)`

`class Vec4D`

`def init(xyzw)`