Outputs
=======

Save geometry and measurements at different stages to inspect an analysis
or resume it without repeating every step.

.. list-table:: Common outputs
   :header-rows: 1
   :widths: 28 44 28

   * - Output
     - Contents
     - How it is produced
   * - ``masks/*_cp_masks.tif``
     - Integer segmentation labels for each selected field.
     - ``segment_images()`` via Omnipose.
   * - ``<dataset>_meshdata.pkl``
     - Cell IDs, field metadata, contours, midlines, and meshes.
     - ``batch_process_mesh(save_data=True)``.
   * - ``<dataset>_curated_meshdata.pkl``
     - Geometry for cells retained by curation.
     - ``curate_dataset(save_curated_data=True)``.
   * - ``<dataset>_meshdata.parquet``
     - Reloadable geometry stored as nested numeric lists.
     - ``meshdata_to_parquet()``.
   * - ``<dataset>_features.pkl`` / ``.parquet``
     - Merged per-cell feature table.
     - ``dataframe_to_pkl()`` / ``dataframe_to_parquet()``.
   * - ``curated_masks/*.tif`` and ``curated_mask_provenance.csv``
     - Rasterized retained contours and original-label overlap records.
     - ``export_curated_masks()``.

The default dataset name is the final component of ``image_folder_path``.
Renaming a working-copy folder therefore changes default output names.
Use explicit filenames when that matters.

Identity columns
----------------

The ordinary table uses ``image_name``, ``cell_id``, and ``frame``. A cell ID is
local to a field; use all identity columns when joining tables. ``frame`` is the
image's collection index and is not evidence of temporal tracking.
``include_metadata_tag=True`` renames these columns to ``Metadata_image_name``,
``Metadata_cell_id``, and ``Metadata_frame``.

See :doc:`../workflow/export` for naming, serialization, overwrite behavior,
and how to preserve processing summaries alongside measurements.