Overview

bioio-conversion

Build Status Documentation PyPI version License Python 3.11–3.13

A BioIO conversion tool for going between image formats.

Documentation

See the full documentation on our GitHub Pages site:

https://bioio-devs.github.io/bioio-conversion

Installation

Install from PyPI along with core BioIO and plugins:

pip install bioio-conversion

Python Package Usage

Available Converters

  • OmeZarrConverter

    • Purpose: Convert any BioImage-supported input (TIFF, CZI, ND2, etc.) into an OME-Zarr store.

    • Features:

      • Multi-scene export (scenes=0, list, or None = all)

      • Flexible multiscale pyramid options (level_shapes, num_levels, downsample_z)

      • Chunk-size tuning (chunk_shape, memory_target, shard_shape)

      • Metadata options (channels, axes_names, axes_units, axes_types, physical_pixel_size)

      • Output format (zarr_format = 2 or 3)

      • Optional auto Dask cluster

    • Import path:

      from bioio_conversion.converters import OmeZarrConverter

  • BatchConverter

    • Purpose: Orchestrate batch conversions of many files (CSV, directory crawl, or explicit list).

    • Features:

      • Factory methods: from_csv(), from_directory(), from_list()

      • Shared default_opts for per-job overrides

      • Dispatch jobs via .run_jobs()

    • Import path:

      from bioio_conversion.converters import BatchConverter

Example: OmeZarrConverter

Minimal usage
from bioio_conversion.converters import OmeZarrConverter

conv = OmeZarrConverter(
    source='image.tiff',
    destination='out_dir'
)
conv.convert()
Advanced usage: full control
from bioio_conversion.converters import OmeZarrConverter
from zarr.codecs import BloscCodec

conv = OmeZarrConverter(
    source='multi_scene.czi',
    destination='zarr_output',
    scenes=None,
    name='experiment1',
    tbatch=2,
    num_levels=3,
    downsample_z=True,
    chunk_shape=(1,1,16,256,256),
    shard_shape=(1,1,128,1024,1024),
    memory_target=32*1024*1024,
    dtype='uint16',
    compressor=BloscCodec(),
    zarr_format=3,
)
conv.convert()
Explicit level_shapes
conv = OmeZarrConverter(
    source="image_tczyx.tif",
    destination="out_tczyx",
    level_shapes=[
        (1, 3, 5, 325, 475),
        (1, 3, 2, 162, 238),
        (1, 3, 1, 81, 119),
    ],
)
conv.convert()
Channel metadata
from bioio_ome_zarr.writers import Channel

channels = [
    Channel(label="DAPI", color="#0000FF", active=True,
            window={"min":100, "max":2000, "start":200, "end":1200}),
    Channel(label="GFP", color="#00FF00", active=True),
    Channel(label="TRITC", color="#FF0000", active=False),
]

conv = OmeZarrConverter(
    source="multi_channel.czi",
    destination="out_channels",
    channels=channels,
)
conv.convert()
Axes & physical pixel sizes
conv = OmeZarrConverter(
    source="custom_axes.tif",
    destination="out_axes",
    axes_names=["t","c","z","y","x"],
    axes_types=["time","channel","space","space","space"],
    axes_units=[None, None, "micrometer","micrometer","micrometer"],
    physical_pixel_size=[1.0, 1.0, 0.4, 0.108, 0.108],
)
conv.convert()
Example with fewer dimensions (3D ZYX)
conv = OmeZarrConverter(
    source="volume_zyx.tif",
    destination="out_zyx",
    num_levels=2,
    downsample_z=True,
)
conv.convert()

CSV-driven batch conversion

The CSV file should have a header row that names the job parameters. At minimum, include a source column (path to each input image). You may also include per-job overrides for any converter option (e.g. destination, scenes, tbatch, num_levels, downsample_z, level_shapes, memory_target, dtype, channel_names, etc.). Values in each row will be merged with the default_opts you passed to BatchConverter.

from bioio_conversion import BatchConverter

bc = BatchConverter(
    converter_key='ome-zarr',
    default_opts={
        'destination': 'batch_out',
        'tbatch': 4,
    }
)
jobs = bc.from_csv('jobs.csv')  # parse CSV into job dicts
bc.run_jobs(jobs)

Directory-driven batch conversion

from bioio_conversion import BatchConverter

bc = BatchConverter(default_opts={
    'destination': 'dir_out',
})
jobs = bc.from_directory(
    '/data/images',
    max_depth=2,
    pattern='*.tif'
)
bc.run_jobs(jobs)

List-driven batch conversion

from bioio_conversion import BatchConverter

paths = ['/data/a.czi', '/data/b.czi', '/data/c.zarr']
bc = BatchConverter(default_opts={
    'destination': 'list_out',
    'scenes': 0
})
jobs = bc.from_list(paths)
bc.run_jobs(jobs)

Command-Line Interface

bioio-convert – single-file conversion

Convert a single image file to OME-Zarr using the configured backend (default: OME-Zarr).

bioio-convert SOURCE -d DESTINATION [options]

SOURCE is the input image file (e.g. .czi, .ome.tiff, .nd2).

Core options

  • source (positional): input image path

  • -d, --destination: output directory for .ome.zarr (required)

  • -n, --name: base name for the output (defaults to a value derived from the input)

  • -s, --scenes: scene(s) to export (e.g. 0 or 0,2). If omitted, the converter/writer default is used (“all scenes”).

  • --tbatch: number of timepoints per write batch.

  • --start-t-src: source T index at which to begin reading (0-based).

  • --start-t-dest: destination T index at which to begin writing (0-based).

Multiscale (pyramid) options

  • --level-shapes: semicolon-separated per-level shapes (level 0 first). Each tuple must have one integer per axis.

    • Example: --level-shapes "1,3,5,512,512;1,3,5,256,256;1,3,5,128,128"

  • --num-levels: total number of pyramid levels (including level 0). If provided (and --level-shapes is not), a half-pyramid is built in X/Y (and optionally Z).

  • --downsample-z: when used with --num-levels, also halves the Z dimension at each level if a Z axis exists.

Chunking / sharding (advanced)

  • --chunk-shape: single chunk shape tuple applied to all levels (e.g. 1,1,16,256,256).

  • --chunk-shape-per-level: semicolon-separated chunk shapes per level. Overrides --chunk-shape and --memory-target.

  • --memory-target: approximate in-memory byte budget used to derive per-level chunk shapes when explicit chunk shapes are not provided.

  • --shard-shape: single shard shape tuple for Zarr v3 (e.g. 1,1,128,1024,1024).

  • --shard-shape-per-level: semicolon-separated shard shapes per level (Zarr v3). Overrides --shard-shape.

Writer / metadata options

  • --dtype: output dtype override (e.g. uint16, float32). If omitted, the reader’s native dtype is used.

  • --physical-pixel-sizes: comma-separated floats (one per axis, level 0 only). Example for (t,c,z,y,x): --physical-pixel-sizes 1.0,1.0,0.4,0.108,0.108

  • --zarr-format: target Zarr version:

    • 2 ≈ NGFF 0.4

    • 3 ≈ NGFF 0.5 If omitted, the writer’s default is used (3 ≈ NGFF 0.5).

Channel display options

These only take effect when --channel-labels is provided. All lists must align by channel index.

  • --channel-labels: comma-separated channel names (e.g. DAPI,GFP,TRITC).

  • --channel-colors: comma-separated colors (CSS color names or hex codes). Example: "#0000FF,#00FF00,#FF0000".

  • --channel-actives: booleans for channel visibility (e.g. true,true,false).

  • --channel-coefficients: per-channel intensity coefficients (e.g. 1,0.8,1.2).

  • --channel-families: intensity family names per channel (e.g. linear,sRGB,sRGB).

  • --channel-inverted: booleans for inverted display per channel.

  • --channel-window-min, --channel-window-max, --channel-window-start, --channel-window-end: per-channel windowing values. Only used when any window value is provided.

Axis metadata options

  • --axes-names: comma-separated axis names in native axis order. Example: t,c,z,y,x.

  • --axes-types: comma-separated axis semantic types (e.g. time,channel,space,space,space).

  • --axes-units: comma-separated axis units, in the same order as --axes-names. Use none, null, or a blank position for missing units. Example for (t,c,z,y,x): s,,um,um,um.

bioio-convert examples

Basic usage

bioio-convert image.tif -d out_dir

Custom name

bioio-convert sample.czi -d out_dir -n my_run

Export all scenes

bioio-convert multi_scene.ome.tiff -d zarr_out

Export specific scenes

bioio-convert multi_scene.ome.tiff -d zarr_out -s 0,2

Simple half-pyramid (XY only)

bioio-convert volume.tif -d out_xy --num-levels 3

Simple half-pyramid (XYZ)

bioio-convert volume_tczyx.tif -d out_xyz --num-levels 3 --downsample-z

Explicit level shapes

bioio-convert image.tif -d out_explicit \
  --level-shapes "1,3,5,325,475;1,3,2,162,238;1,3,1,81,119"

Dtype and chunking

bioio-convert image.tif -d out_dir \
  --dtype uint16 \
  --memory-target 33554432

Custom channels

bioio-convert image_with_channels.czi -d out_dir \
  --channel-labels DAPI,GFP,TRITC \
  --channel-colors "#0000FF,#00FF00,#FF0000" \
  --channel-actives true,true,false

Axis metadata

bioio-convert image_tczyx.tif -d out_axes \
  --axes-names t,c,z,y,x \
  --axes-types time,channel,space,space,space \
  --axes-units s,,um,um,um

Physical pixel sizes

bioio-convert image.tif -d out_dir \
  --physical-pixel-sizes 1.0,1.0,0.4,0.108,0.108

bioio-batch-convert – batch conversion

Batch mode: convert many files via CSV, directory walk, or an explicit list of paths. All of the shared OME-Zarr options listed for bioio-convert (--num-levels, --chunk-shape, --channel-*, axis metadata, etc.) are also accepted here and act as defaults for every job.

bioio-batch-convert --mode [csv|dir|list] [mode options] [shared options]

Mode selection

  • -m, --mode [csv|dir|list] (required):

    • csv: read jobs from a CSV file.

    • dir: scan a directory tree for input files.

    • list: use an explicit list of paths from the command line.

Mode-specific options

  • CSV mode (--mode csv)

    • --csv-file: path to a CSV describing jobs (one row per job). Each column name maps to an OmeZarrConverter init argument (e.g. source, destination, scenes, tbatch, etc.). Values are parsed by the batch loader; per-row values override shared defaults from the CLI.

  • Directory mode (--mode dir)

    • --directory / --dir: root directory to scan.

    • --depth: maximum recursion depth (0 = only top-level files).

    • --pattern: glob pattern used when scanning (e.g. “*.czi”).

  • List mode (--mode list)

    • --paths: explicit input file paths (repeatable).

Shared conversion options

All of the bioio-convert options (destination, multiscale, chunking, channels, axes, etc.) can be passed to bioio-batch-convert. They are converted via build_ome_zarr_init_opts(...) and applied as defaults to every job created by the BatchConverter. CSV columns that match a given argument override the shared defaults on a per-job basis.

bioio-batch-convert examples

CSV mode

bioio-batch-convert \
  --mode csv \
  --csv-file jobs.csv \
  --destination batch_out \
  --tbatch 4 \
  --dtype uint16 \
  --num-levels 3

Directory mode

bioio-batch-convert \
  --mode dir \
  --directory data/ \
  --depth 2 \
  --pattern '*.czi' \
  --destination output_zarr \
  --level-shapes "1,3,5,325,475;1,3,2,162,238;1,3,1,81,119"

List mode

bioio-batch-convert \
  --mode list \
  --paths a.czi b.czi c.tiff \
  --destination list_out \
  --name batch_run \
  --num-levels 2 \
  --downsample-z

License & Issues

BSD 3-Clause https://bioio-devs.github.io/bioio-conversion/LICENSE

Report bugs at: https://github.com/bioio-devs/bioio-conversion/issues