Sensor data labeling

Labeling for robotics and autonomy data

Avala labels camera, LiDAR, point-cloud, and sequence data for robotics, then keeps that work inside the fleet learning loop: ingest, review, a versioned dataset, and the training handoff.

Fleet Learning Infrastructure for Physical AI

Published robotics labeling capabilities

Published robotics labeling capabilities
CapabilityStatusWhat Avala publishes
Cameras and videoSupportedImages and video are labeled with 2D boxes and masks, alongside the sequence they belong to.
LiDAR and point cloudsSupportedPoint-cloud scenes take 3D cuboids and point-cloud labels, with camera and sequence context kept where the workflow supports them.
Depth mapsSupportedDepth maps are a published annotation type on the same workflow as video, point clouds, text, and audio.
MCAP, ROS 1, and ROS 2SupportedMCAP playback follows supported message schemas. The ROS bag SDK importer converts cameras and supported numeric/text topics; it reports skipped topics, including point clouds.Supported message types
Timestamp and calibration contextPreservedTimestamp and calibration context stay attached while teams inspect the same event across sensor streams.
2D boxes, masks, polylines, and classificationsSupportedThese are published ground-truth outputs for camera and sequence data.
3D cuboidsSupportedCuboids capture position, dimensions, and heading in point-cloud scenes.
Tracking across a sequenceSupportedTracked objects are a published output, so an object keeps its identity across the sequence.
Trajectory contextSupportedCamera, point-cloud, and trajectory data stay aligned as work moves through labeling and review.
LeRobot datasetsSupportedImport LeRobot data as MCAP recordings. The LeRobot v3 round trip preserves numeric and text columns; camera frames are re-encoded, so pixels are not identical.

One Avala sequence becomes one LeRobot episode. State and actions require numeric source fields. Optional outcome labels are written to a sidecar; sequence-export timestamps follow the configured frame rate.

LeRobot conversion guide

What reaches your training loader?

Inspect the field mappings for two LeRobot v3 export paths. Required inputs differ: camera data alone does not provide robot actions. Expand either path to check its requirements before choosing a converter.

Avala sequences → LeRobot v3

Default exports contain cameras, timestamps and task text. State and action require configured numeric source fields; missing values are errors. Outcomes are optional sidecar metadata. Sequence timestamps use the configured frame rate.

Export field reference
observation.images.*
--state-key → observation.state
--action-key → action
frame_index / fps → timestamp
--include-outcomes → meta/avala_sequence_outcomes.jsonl
Robot MCAP → LeRobot v3

Joint states provide measured positions; a separate command topic provides actions. Frames follow the first camera, with other topics sampled at or before each frame. The export report records timing deviation, dropped leading frames and video encoding mode.

Export field reference
foxglove.CompressedImage / CompressedVideo → observation.images.*
/joint_states → observation.state
/joint_commands → action
frame_index / fps → timestamp
meta/avala_export_report.json
LeRobot conversion guide

How should you evaluate a data program?

Use this checklist to scope a benchmark with Avala or another provider. Agree on evidence and acceptance criteria before work starts. Checking a box records your review in this page only; it does not certify a provider or submit a request.

Blank planning templates in English. Define requirements and record evidence before agreeing on a program. These files contain no results or commercial terms.

Questions this page answers

What sensor data can Avala label for robotics?

Avala publishes workflows for cameras, video, LiDAR, point clouds, depth maps, and sequences, and it imports MCAP, ROS 1, and ROS 2 recordings with synchronized camera, LiDAR, and telemetry streams.

Is labeling the whole product?

Labeling connects to ingest, visualization, quality review and versioned dataset delivery in Avala's fleet-learning workflow.

How is labeling reviewed?

Labeling runs with human review and traceable quality-control states. The annotation pages describe the review path for camera, LiDAR, and telemetry.

How do I start a robotics labeling workflow?

Book a demo, or read the ground-truth outputs and the MCAP annotation page. LeRobot import and export are documented with the SDKs.