google-deepmind/open_x_embodiment resource

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

2005 stars · 123 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

Open X-Embodiment is a unified collection of open-sourced robot learning datasets from multiple institutions, formatted in the RLDS episode format for downstream training. It also provides RT-X model checkpoints (TensorFlow and JAX) trained on this data for robotic manipulation tasks.

Use cases

• download large-scale robot manipulation datasets in a unified format
• train robot policies on multi-embodiment data
• run inference with pretrained RT-1-X checkpoints
• visualize robot episodes for research
• benchmark robotic learning models
• fine-tune vision-language-action models on robot data

When to choose

• you need diverse, cross-embodiment robot data for training or evaluation
• you want pretrained RT-X checkpoints for robot manipulation research
• you work with RLDS/TensorFlow Datasets pipelines

When to avoid

• you need real-time robot control beyond the supported observation/action spaces
• you need wrist or depth camera inputs, which the provided models do not use
• you want a general-purpose robotics simulator rather than datasets

Facets

dataset · maturity active

machine-learning deep-learning data-science robotics machine-learning deep-learning python cloud robotics-dataset rlds rt-x-models imitation-learning tensorflow-datasets robot-manipulation gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="google-deepmind/open_x_embodiment")