MPI IK

A simple inverse kinematics solver for MANO hand model, SMPL body model, and SMPL-H body+hand model.

Installation

pip install eden-mpi-ik

For development:

pip install "eden-mpi-ik[dev]"

Usage

from mpi_ik import MANO_LEFT_MODEL_PATH, KinematicModel, KinematicPCAWrapper, MANOArmature, Solver

mesh = KinematicModel(MANO_LEFT_MODEL_PATH, MANOArmature, scale=1000)
wrapper = KinematicPCAWrapper(mesh, n_pose=12)
solver = Solver(verbose=True)

_, keypoints = mesh.set_params(pose_pca=pose_pca, pose_glb=pose_glb, shape=shape)
# keypoints (21, 3)
params_est = solver.solve(wrapper, keypoints)

See examples/mano_example.py for a full working example.

Model Download

Download the official model files from the following sites (registration required):

• MANO: https://mano.is.tue.mpg.de/
• SMPL: https://smpl.is.tue.mpg.de/
• SMPL-H: https://mano.is.tue.mpg.de/ (included with MANO download)

Model Preparation

Convert official MANO/SMPL/SMPL-H model files into the format expected by this package. By default, converted models are saved to ~/.cache/mpi_ik/:

# Output defaults to ~/.cache/mpi_ik/mano_left.pkl
mpi-ik-prepare --model-type mano-left --input path/to/mano_v1_2/models/MANO_LEFT.pkl

# Or specify a custom output path
mpi-ik-prepare --model-type mano-left --input path/to/mano_v1_2/models/MANO_LEFT.pkl --output custom/path/mano_left.pkl

The cache directory can be overridden with the MPI_IK_CACHE_DIR environment variable.

Acknowledgement

This project is based on Minimal-IK by Yuxiao Zhou.