Export

How a Stera session lands on disk — what formats, what fields, and where to find the full reference.

Export is the last step of the pipeline. After Process — and ideally a quick QC pass through Evaluate — one call bundles a session into a self-contained directory that drops straight into a training pipeline.

Evaluate(session).show()                      # QC the run (optional)
session.export("episodes/run_01", visualizer=viz)

That one call produces an episode:

episodes/run_01/
├── rgb.mp4              ─ original RGB video
├── mesh.ply             ─ scene mesh
├── thumbnail.jpg
├── annotation.hdf5      ─ all time-series annotations
├── visualization.rrd    ─ Rerun replay file (optional)
└── calibrations/        ─ intrinsics, extrinsics, depth↔RGB transforms

What's inside

Output formats

MCAP in, HDF5 + PLY + MP4 out. The full field-level reference lives in Process > Concepts.

Episode export guide

The full guide for session.export(), including options and the returned manifest.

HDF5 schema

Every dataset, dtype, and shape inside annotation.hdf5.

Evaluate the recording

Generate an HTML report with metrics + a health score before bundling the episode.

Calibrations, coordinate frames, and what each pose actually means is documented in Process > Concepts. Export is the packaging step — semantics live there.