mocap-converter 0.0.1

Motion capture conversion utilities.

mocap-converter

mocap-converter is a library for turning motion-capture recordings into formats that downstream tools can reuse. It converts joint position traces captured during a mocap session into joint rotation data, making it easy to repurpose captured motion in other contexts.

Installation

Install with uv:

$ uv add mocap-converter

Or install with pip:

$ pip install mocap-converter

Usage

Convert Azure Kinect Data to BVH

The snippet below loads Azure Kinect Body Tracking data exported by the Offline Processor, derives joint rotations from the recorded global joint positions, and saves the result as a BVH animation.

import json

from mocap_converter.adapter.azure_kinect import (
    AZURE_KINECT_KINEMATIC_TREE,
    get_positions_from_json,
)
from mocap_converter.adjust_kinematic_tree import adjust_kinematic_tree
from mocap_converter.io.bvh.saver import save_bvh
from mocap_converter.motion_data import MotionData
from mocap_converter.pos2rot import get_rotations_from_positions


with open("path/to/kinect_capture.json") as raw:
    kinect_data = json.load(raw)

positions = get_positions_from_json(kinect_data)
# Apply optional smoothing (e.g., a median filter) here if the capture contains noisy frames.
kinematic_tree = adjust_kinematic_tree(AZURE_KINECT_KINEMATIC_TREE, positions)
root = kinematic_tree.root

positional_motion = MotionData(kinematic_tree, positions)
rotations = get_rotations_from_positions(positional_motion, current_node_name=root.name)

rotational_motion = MotionData(
    kinematic_tree,
    positions={root.name: positions[root.name]},
    rotations=rotations,
)

save_bvh(rotational_motion, "path/to/output.bvh")

License

mocap-converter is distributed under the terms of the MIT license.