retarget

retarget is a typed research toolkit for processing heterogeneous capture and retargeting the resulting scene observation to robots. Native recordings, alignment, semantic contacts, robot adaptation, optimization, export, and visualization use one public abstraction hierarchy.

First time here?

Start with Quickstart, then read Introduction for the complete capture-to-retarget flow. Architecture explains the domain boundaries.

Live code (local only)

Runnable blocks require local uv run mkdocs serve. Hosted documentation is read-only.

CLIPython

uv sync --extra dev
uv run retarget doctor
uv run retarget run --config examples/basic/run_config.toml

from retarget import RetargetingExperiment

result = RetargetingExperiment(
    observation=observation_recipe,
    recipe=retargeting_recipe,
    robot=robot,
).run()

Run configs deserialize into the same observation recipe, retargeting recipe, robot spec, and RetargetingExperiment; they are not a second workflow.

Where To Go Next

Goal	Page
Run a small experiment	Quickstart
Process heterogeneous sensors	Introduction
Understand the abstraction hierarchy	Architecture
Follow the capture-to-retarget workflow	Capture to retarget
Work with frames and clocks	Coordinate Conventions
Extend formats, robots, or recipes	Extending
Browse the API	API