SMPL-H

SMPL-H extends SMPL with articulated MANO hands.

Setup

SMPL-H requires registration at https://mano.is.tue.mpg.de/.

# Download SMPL-H after configuring credentials for the upstream site.
body-models download smplh

Manual paths can also be configured per gender:

# Configure local SMPL-H files when you already have the AMASS models on disk.
body-models set smplh-neutral /path/to/smplh/neutral/model.npz
body-models set smplh-male /path/to/smplh/male/model.npz
body-models set smplh-female /path/to/smplh/female/model.npz

Notes

The downloader uses the "Extended SMPL+H model (used in AMASS project)" archive.

API

body_models.bodies.smplh.numpy.SMPLH

SMPLH(
    model_path=None,
    gender=None,
    flat_hand_mean=False,
    simplify=1.0,
    rotation_type="axis_angle",
    kernel="numpy",
)

Bases: BodyModel

SMPL-H body model with NumPy backend.

Initialize the SMPLH model.

PARAMETER	DESCRIPTION
model_path	Path to model assets, or the default assets when omitted. TYPE: Path | str | None DEFAULT: None
gender	Model gender variant to load. TYPE: Literal['neutral', 'male', 'female'] | None DEFAULT: None
flat_hand_mean	Whether to use a flat hand as the pose mean. TYPE: bool DEFAULT: False
simplify	Mesh simplification factor to apply while loading. TYPE: float DEFAULT: 1.0
rotation_type	Rotation representation expected by pose inputs. TYPE: RotationType DEFAULT: 'axis_angle'
kernel	Backend kernel used for forward evaluation. TYPE: Literal['numpy', 'scipy', 'numba'] DEFAULT: 'numpy'

METHOD	DESCRIPTION
forward_vertices	Compute posed mesh vertices.
forward_skeleton	Compute posed joint transforms.
prepare_identity	Precompute shape-dependent state for repeated forward passes.
prepare_pose	Precompute pose-dependent state for repeated forward passes.
joint_index	Resolve a standard joint to this model's native joint index.
prepare_skinning	Pack prepared model state into renderer-ready skinning inputs.

ATTRIBUTE	DESCRIPTION
common_joints	Common anatomical joints mapped to this model's native joint names. TYPE: Mapping[Joint, str]

Source code in src/body_models/bodies/smplh/numpy.py

def __init__(
    self,
    model_path: Path | str | None = None,
    gender: Literal["neutral", "male", "female"] | None = None,
    flat_hand_mean: bool = False,
    simplify: float = 1.0,
    rotation_type: RotationType = "axis_angle",
    kernel: Literal["numpy", "scipy", "numba"] = "numpy",
):
    """Initialize the SMPLH model.

    Args:
        model_path: Path to model assets, or the default assets when omitted.
        gender: Model gender variant to load.
        flat_hand_mean: Whether to use a flat hand as the pose mean.
        simplify: Mesh simplification factor to apply while loading.
        rotation_type: Rotation representation expected by pose inputs.
        kernel: Backend kernel used for forward evaluation.
    """
    if gender is not None and gender not in ("neutral", "male", "female"):
        raise ValueError(f"Invalid gender: {gender}. Must be 'neutral', 'male', or 'female'.")
    if rotation_type not in VALID_ROTATION_TYPES:
        raise ValueError(f"Invalid rotation_type: {rotation_type}")
    if simplify < 1.0:
        raise ValueError("simplify must be >= 1.0")
    if kernel not in self.kernels:
        raise ValueError(f"Invalid kernel: {kernel}")

    self.gender = gender if gender is not None else "neutral"
    self.rotation_type = rotation_type
    self.num_rot_dims = 2 if rotation_type in ("matrix", "rotmat") else 1
    self._kernel = _get_kernel(kernel)

    resolved_path = get_model_path(model_path, gender)
    self.weights = load_model_data(resolved_path, flat_hand_mean=flat_hand_mean, simplify=simplify)

common_joints property

common_joints

Common anatomical joints mapped to this model's native joint names.

forward_vertices

forward_vertices(
    body_pose,
    hand_pose,
    pelvis_rotation=None,
    global_rotation=None,
    global_translation=None,
    vertex_indices=None,
    *,
    shape=None,
    identity=None,
)

Compute posed mesh vertices.

PARAMETER	DESCRIPTION
shape	Shape coefficients. TYPE: Float[ndarray, '*batch 10'] | None DEFAULT: None
body_pose	Local body joint rotations. TYPE: Float[ndarray, '*batch 21 N'] | Float[ndarray, '*batch 21 3 3']
hand_pose	Local hand joint rotations. TYPE: Float[ndarray, '*batch 30 N'] | Float[ndarray, '*batch 30 3 3']
pelvis_rotation	Root pelvis rotation. TYPE: Float[ndarray, '*batch N'] | Float[ndarray, '*batch 3 3'] | None DEFAULT: None
global_rotation	Global model rotation. TYPE: Float[ndarray, '*batch N'] | Float[ndarray, '*batch 3 3'] | None DEFAULT: None
global_translation	Global model translation. TYPE: Float[ndarray, '*batch 3'] | None DEFAULT: None
vertex_indices	Optional subset of vertices to return. TYPE: Any | None DEFAULT: None

RETURNS	DESCRIPTION
Float[ndarray, '*batch V 3']	Posed vertex positions.

Source code in src/body_models/bodies/smplh/numpy.py

def forward_vertices(
    self,
    body_pose: Float[np.ndarray, "*batch 21 N"] | Float[np.ndarray, "*batch 21 3 3"],
    hand_pose: Float[np.ndarray, "*batch 30 N"] | Float[np.ndarray, "*batch 30 3 3"],
    pelvis_rotation: Float[np.ndarray, "*batch N"] | Float[np.ndarray, "*batch 3 3"] | None = None,
    global_rotation: Float[np.ndarray, "*batch N"] | Float[np.ndarray, "*batch 3 3"] | None = None,
    global_translation: Float[np.ndarray, "*batch 3"] | None = None,
    vertex_indices: Any | None = None,
    *,
    shape: Float[np.ndarray, "*batch 10"] | None = None,
    identity: SmplhIdentity | None = None,
) -> Float[np.ndarray, "*batch V 3"]:
    """Compute posed mesh vertices.

    Args:
        shape: Shape coefficients.
        body_pose: Local body joint rotations.
        hand_pose: Local hand joint rotations.
        pelvis_rotation: Root pelvis rotation.
        global_rotation: Global model rotation.
        global_translation: Global model translation.
        vertex_indices: Optional subset of vertices to return.

    Returns:
        Posed vertex positions.
    """
    if identity is None:
        assert shape is not None
        batch_shape = body_pose.shape[: -(self.num_rot_dims + 1)]
        shape = np.broadcast_to(shape, (*batch_shape, shape.shape[-1]))
        identity = self.prepare_identity(shape)
    pose = self.prepare_pose(body_pose, hand_pose, pelvis_rotation, identity=identity)
    return self._kernel.forward_vertices(
        self.weights,
        identity["rest_vertices"],
        pose["skinning_transforms"],
        pose["pose_offsets"],
        global_rotation=global_rotation,
        global_translation=global_translation,
        vertex_indices=vertex_indices,
        rotation_type=self.rotation_type,
    )

forward_skeleton

forward_skeleton(
    body_pose,
    hand_pose,
    pelvis_rotation=None,
    global_rotation=None,
    global_translation=None,
    joint_indices=None,
    *,
    shape=None,
    identity=None,
)

Compute posed joint transforms.

PARAMETER	DESCRIPTION
shape	Shape coefficients. TYPE: Float[ndarray, '*batch 10'] | None DEFAULT: None
body_pose	Local body joint rotations. TYPE: Float[ndarray, '*batch 21 N'] | Float[ndarray, '*batch 21 3 3']
hand_pose	Local hand joint rotations. TYPE: Float[ndarray, '*batch 30 N'] | Float[ndarray, '*batch 30 3 3']
pelvis_rotation	Root pelvis rotation. TYPE: Float[ndarray, '*batch N'] | Float[ndarray, '*batch 3 3'] | None DEFAULT: None
global_rotation	Global model rotation. TYPE: Float[ndarray, '*batch N'] | Float[ndarray, '*batch 3 3'] | None DEFAULT: None
global_translation	Global model translation. TYPE: Float[ndarray, '*batch 3'] | None DEFAULT: None
joint_indices	Optional subset of joints to return. TYPE: Any | None DEFAULT: None

RETURNS	DESCRIPTION
Float[ndarray, '*batch 52 4 4']	Joint transforms in the model hierarchy.

Source code in src/body_models/bodies/smplh/numpy.py

def forward_skeleton(
    self,
    body_pose: Float[np.ndarray, "*batch 21 N"] | Float[np.ndarray, "*batch 21 3 3"],
    hand_pose: Float[np.ndarray, "*batch 30 N"] | Float[np.ndarray, "*batch 30 3 3"],
    pelvis_rotation: Float[np.ndarray, "*batch N"] | Float[np.ndarray, "*batch 3 3"] | None = None,
    global_rotation: Float[np.ndarray, "*batch N"] | Float[np.ndarray, "*batch 3 3"] | None = None,
    global_translation: Float[np.ndarray, "*batch 3"] | None = None,
    joint_indices: Any | None = None,
    *,
    shape: Float[np.ndarray, "*batch 10"] | None = None,
    identity: SmplhIdentity | None = None,
) -> Float[np.ndarray, "*batch 52 4 4"]:
    """Compute posed joint transforms.

    Args:
        shape: Shape coefficients.
        body_pose: Local body joint rotations.
        hand_pose: Local hand joint rotations.
        pelvis_rotation: Root pelvis rotation.
        global_rotation: Global model rotation.
        global_translation: Global model translation.
        joint_indices: Optional subset of joints to return.

    Returns:
        Joint transforms in the model hierarchy.
    """
    if identity is None:
        assert shape is not None
        batch_shape = body_pose.shape[: -(self.num_rot_dims + 1)]
        shape = np.broadcast_to(shape, (*batch_shape, shape.shape[-1]))
        identity = self.prepare_identity(shape, skip_vertices=True)
    pose = self.prepare_pose(body_pose, hand_pose, pelvis_rotation, identity=identity, skip_vertices=True)
    return self._kernel.forward_skeleton(
        self.weights,
        pose["skeleton_transforms"],
        global_rotation=global_rotation,
        global_translation=global_translation,
        joint_indices=joint_indices,
        rotation_type=self.rotation_type,
    )

prepare_identity

prepare_identity(shape, skip_vertices=False)

Precompute shape-dependent state for repeated forward passes.

Source code in src/body_models/bodies/smplh/numpy.py

def prepare_identity(
    self,
    shape: Float[np.ndarray, "*batch 10"],
    skip_vertices: bool = False,
) -> SmplhIdentity:
    """Precompute shape-dependent state for repeated forward passes."""
    return self._kernel.prepare_identity(self.weights, shape, skip_vertices=skip_vertices)

prepare_pose

prepare_pose(
    body_pose,
    hand_pose,
    pelvis_rotation=None,
    *,
    identity,
    skip_vertices=False,
)

Precompute pose-dependent state for repeated forward passes.

Source code in src/body_models/bodies/smplh/numpy.py

def prepare_pose(
    self,
    body_pose: Float[np.ndarray, "*batch 21 N"] | Float[np.ndarray, "*batch 21 3 3"],
    hand_pose: Float[np.ndarray, "*batch 30 N"] | Float[np.ndarray, "*batch 30 3 3"],
    pelvis_rotation: Float[np.ndarray, "*batch N"] | Float[np.ndarray, "*batch 3 3"] | None = None,
    *,
    identity: SmplhIdentity,
    skip_vertices: bool = False,
) -> SmplhPreparedPose:
    """Precompute pose-dependent state for repeated forward passes."""
    return self._kernel.prepare_pose(
        self.weights,
        body_pose,
        hand_pose,
        pelvis_rotation,
        rotation_type=self.rotation_type,
        local_joint_offsets=identity["local_joint_offsets"],
        rest_joints=identity["rest_joints"],
        skip_vertices=skip_vertices,
    )

joint_index

joint_index(joint)

Resolve a standard joint to this model's native joint index.

Source code in src/body_models/base.py

def joint_index(self, joint: Joint) -> int:
    """Resolve a standard joint to this model's native joint index."""
    if not isinstance(joint, Joint):
        raise TypeError("joint_index() expects a body_models.Joint; use joint_names.index(...) for native names.")
    try:
        native_name = self.common_joints[joint]
    except KeyError as exc:
        raise KeyError(f"{self.__class__.__name__} has no standard joint {joint.value!r}") from exc
    return self.joint_names.index(native_name)

prepare_skinning

prepare_skinning(*, identity, pose)

Pack prepared model state into renderer-ready skinning inputs.

Source code in src/body_models/base.py

def prepare_skinning(self, *, identity: Mapping[str, Any], pose: Mapping[str, Any]) -> SkinningPayload:
    """Pack prepared model state into renderer-ready skinning inputs."""
    if self.is_rigid_body:
        raise NotImplementedError(f"{self.__class__.__name__} is rigid and does not support skinning.")

    skinning: SkinningPayload = {
        "rest_vertices": identity["rest_vertices"],
        "skinning_transforms": pose["skinning_transforms"],
        "skin_weights": self.skin_weights,
        "faces": self.faces,
    }
    if "pose_offsets" in pose:
        skinning["pose_offsets"] = pose["pose_offsets"]
    return skinning