fit_with_known_pose#

smplfitter.np.BodyFitter.fit_with_known_pose(pose_rotvecs, target_vertices, target_joints=None, vertex_weights=None, joint_weights=None, beta_regularizer=1.0, beta_regularizer2=0.0, scale_regularizer=0.0, kid_regularizer=None, share_beta=False, scale_target=False, scale_fit=False, beta_regularizer_reference=None, kid_regularizer_reference=None, requested_keys=('shape_betas',))[source]#

Fits the body shape and translation with known output pose.

Parameters:

  • pose_rotvecs – The known output joint rotations as rotation vectors, shaped as (batch_size, num_joints * 3).

  • target_vertices – Target mesh vertices, shaped as (batch_size, num_vertices, 3).

  • target_joints – Optional target joint positions.

  • vertex_weights – Optional importance weights for vertices.

  • joint_weights – Optional importance weights for joints.

  • beta_regularizer – L2 regularization weight for shape parameters.

  • beta_regularizer2 – Secondary regularization for first two shape params.

  • scale_regularizer – Regularization for scale factor.

  • kid_regularizer – Regularization for kid blendshape factor.

  • share_beta – Whether to share shape params across batch.

  • scale_target – Whether to estimate scale for target vertices.

  • scale_fit – Whether to estimate scale for fitted mesh.

  • beta_regularizer_reference – Reference values for beta regularization.

  • kid_regularizer_reference – Reference values for kid factor regularization.

  • requested_keys – List of result keys to return.

Returns:

Dictionary with shape_betas, trans, and optionally kid_factor, scale_corr.