ABC-130k — Rerun RRD conversion

Per-episode Rerun recordings converted from the XDOF/ABC-130k robot-teleoperation dataset. Each episode is one .rrd file at base/<task>__<uuid>.rrd, ready to view in the Rerun Viewer or to register as a catalog dataset. The default viewer layout can be found at blueprints/default.rbl.

Provenance and attribution

This dataset is a converted derivative of XDOF/ABC-130k, released by its original authors under the Apache License 2.0 (source revision 29136bc9, dated 2026-07-02). The source dataset is gated: to browse it or fetch the originals, accept the terms on its Hugging Face page. All credit for the underlying data collection belongs to the original authors. The original authors do not endorse this conversion.

Changes from the source

• Each episode's episode.mcap (plus annotation.mcap when present) is converted into one Rerun .rrd recording.
• 1920×1200 streams are downscaled to 640×400 with the camera intrinsics rescaled to match.
• Camera video is re-encoded for finer seeking.

Note: This dataset version is larger than the original due to re-encoded videos. Our conversion code provides options for transcoding, including no re-encoding. The table below compares the file sizes for several sample episodes. The first episode is the only one here whose cameras are 1920×1200, so it is the only one the downscale applies to.

Contents

Each recording carries the source episode's streams on their MCAP-native timelines:

• Camera video streams with pinhole calibration.
• Per-joint arm and gripper state/action series (position, velocity, torque) for both arms.
• The task instruction, plus timestamped subtask labels for annotated episodes.

130,703 episodes, ~33 TB total.

How to use

For a minimal example, fetch an episode from the bucket and view it locally:

pip install "huggingface_hub[cli]" rerun-sdk
hf buckets cp hf://buckets/rerun/abc-130k/base/<task>__<uuid>.rrd .
hf buckets cp hf://buckets/rerun/abc-130k/blueprints/default.rbl .
rerun <task>__<uuid>.rrd default.rbl

The converted recordings were written with Rerun 0.34.1. A newer viewer reads them as they are in most cases; rerun rrd migrate updates them if it cannot.

To copy the whole bucket, mirror it bucket-to-bucket rather than to a local disk. The bucket is reachable over an S3-compatible API, so rclone can stream it straight into your own S3-compatible storage.

Scope

Conversion code is published at rerun-io/rrd-datasets. The code and these recordings demonstrate the conversion; they are not a versioned dataset release. Files may be regenerated or replaced, and converting the same episode twice does not give identical bytes.

License

Apache License 2.0, the same license as the source dataset — the full text ships as the LICENSE file at the root of this bucket. If you redistribute or build on this data, keep the attribution above and comply with the license terms.