dex-retargeting

What it is and what it does

Dex Retargeting is a Python library that maps human hand motion to robot hand motion through a suite of retargeting optimizers. It originated from the AnyTeleop project and supports two main workflows: translating human hand video into robot commands (useful for teleoperation systems) and post-processing hand-object pose datasets for robot imitation learning. The library handles the complex problem of mapping human hand kinematics to different robot hand morphologies, accounting for differences in joint counts, ranges, and configurations.

The package depends on anytree, lxml, nlopt, numpy, pin, pytransform3d, and pyyaml to perform optimization and geometric transformations. It explicitly manages joint ordering through named joint identifiers to ensure compatibility with external simulators and robot drivers. The library is in alpha status and has not seen a release in over a year, though the repository remains active.

Use it for:

• Teleoperation systems that stream human hand video and need robot hand control commands.
• Post-processing human demonstration datasets to retarget poses for robot imitation learning tasks.
• Mapping hand-object interaction data from human demonstrations to dexterous robot platforms with different morphologies.
• Research on dexterous manipulation where human motion capture needs translation to specific robot hand models.
• Integration with simulators for validation and testing of retargeted hand trajectories.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Translates human hand motion to robot hand motion through retargeting optimizers, supporting both video-based and pose-dataset-based workflows.

Yes, with conditions. Install if you need to map human hand motion to a dexterous robot hand and are comfortable with alpha-stage software. The low install friction and permissive MIT license are favorable. However, the aging maintenance status and reliance on compiled dependencies like nlopt mean you should verify platform compatibility and be prepared to troubleshoot if issues arise. No known security vulnerabilities.

Install

dex-retargeting on PyPI

pip install dex-retargeting

uv add dex-retargeting

poetry add dex-retargeting

Installing dex-retargeting

Before you install

Low friction install with 7 runtime dependencies. Maintenance status is aging—last release was 496 days ago—but the repository remains active with recent commits and no archived status.

License in practice

MIT license is permissive, allowing commercial and private use with minimal restrictions. You may use, modify, and distribute this package freely provided you include the original license notice.

Quickstart

pip install dex_retargeting

from dex_retargeting.seq_retarget import SeqRetargeting
import numpy as np

retargeting = SeqRetargeting()
retarget_qpos = retargeting.retarget(human_hand_pose)

Requires URDF robot hand model files; joint ordering must be handled explicitly by joint names when integrating with other simulators or robot drivers.

Verify before relying

• Whether the 7 runtime dependencies (anytree, lxml, nlopt, numpy, pin, pytransform3d, pyyaml) are all necessary for core retargeting or only for optional features.
• Performance characteristics and accuracy metrics for different retargeting optimizer types under typical use conditions.
• Whether nlopt requires system-level build tools or if pre-built wheels are available for all supported platforms.
• Specific Python version compatibility within the 3.7–3.10 range and any known issues with numpy >= 2.0.0 beyond the changelog note.

Package facts

Evidence: dex_retargeting-0.5.0-py3-none-any.whl

Tags

More Artificial Intelligence packages

Further reading