Commands

--version flag

Show the version and exit.

--raw

The image to segment: a Zarr array or container, a tif, or a folder of 2D images. A glob makes one volume per match. Example: --raw data.zarr/raw/s0

--voxel-size

The raw's voxel size in nm, z y x, when its metadata has none. Example: --voxel-size 50,4,4

--labels

The traced objects to train on, as a path relative to the raw's container or an absolute path. Without labels, training does not use the volume. Example: --labels labels/s0

--labelled

How the labels cover the raw: sparsely, densely, or a mask path (densely labelled within it). Example: --labelled sparsely

--untraced

With sparse labels, what fills the untraced space next to a traced object: other (objects) or background. Example: --untraced background

--target-mask

Segment, refine and evaluate only inside this mask; predict covers the mask array's extent. Example: --target-mask mask/s0

--gt

The ground truth for the evaluate step; 'labels' reuses the labels. bs prepare adds an evaluate step if any volume has ground truth. Example: --gt labels

--name

The volume's name, used in its table names, such as [segment.apical]. The default is the container's file name, such as data-zarr. Example: --name apical

--models

The models in run order, by name or short form; the default fits the labels. Example: --models 2M,3Af2L

--only text, repeatable

Run only these steps: a kind, kind.name, or a file prefix in a directory

--from text

Run from this step to the end

--dry-run flag

Check the input paths and print what each selected step reads and writes, without running anything

--snapshot / -s directory

Path to the Zarr container of a snapshot

--bind / -b text; default 0.0.0.0

The address to serve the viewer on

--step / -s text

The train step to run when the file has several

--max-iterations / -n integer

Number of training iterations

--save-checkpoints-every / -ce integer

Save checkpoints every n iterations

--save-snapshots-every integer

Save snapshots every n iterations

--voxel-size / -v text

Voxel size in nm, z y x (space-separated integers)

--step / -s text

The predict step to run when the file has several

--roi-offset / -ro text

ROI offset in world units, z y x (space-separated integers)

--roi-shape / -rs text

ROI shape in world units, z y x (space-separated integers)

--num-workers / -w integer

Number of workers

--num-gpus / -ng integer

Number of GPUs to use

--step text

The segment step to run when the file has several

--ws / -ws flag

Watershed segmentation (waterz)

--mws / -mws flag

Mutex watershed segmentation

--cc / -cc flag

Connected components segmentation

--roi-offset / -ro text

ROI offset in world units, z y x (space-separated integers)

--roi-shape / -rs text

ROI shape in world units, z y x (space-separated integers)

--blockwise / -b flag

Run ws or mws over many daisy blocks. Without this flag, the blockwise key of the config applies, and an unset key means one block. cc cannot run blockwise.

--num-workers / -w integer

Number of workers, for blockwise segmentation

--block-shape / -bs text

Block shape in voxels, z y x (space-separated integers). Use 'roi' for one block.

--block-context / -bc text

Block context in voxels, z y x, for blockwise segmentation (space-separated integers)

--param / -p text, repeatable

A method parameter that overrides the config, as key=value. Example: -p 'thresholds=[0.2,0.3]'

--in-array / -i path, required

The label array to read.

--mask / -m path, required

The mask array, at the labels' voxel size. Labels become 0 where the mask is 0.

--out-array / -o path

The label array to write. Default: the input path with the suffix _masked.

--num-workers / -w integer; default 20

The number of daisy workers that process the blocks.

--in-array / -i path, required

The label array to read.

--out-array / -o path

The label array to write. Default: append the operation to the input path, such as labels_fill_holes.

--op dilate or erode or opening or closing or fill-holes, required

The morphological operation. fill-holes gives each region that one object encloses that object's id. It always fills background; it absorbs enclosed objects when the region has at most --max-area voxels.

--iterations / -n integer; default 1

The number of iterations for dilate, erode, opening and closing.

--max-area integer; default 0

The largest enclosed region in which fill-holes absorbs objects, in voxels (pixels with --xy), such as 84000. A value of 0 keeps every object.

--exclude-ids text

Label ids that dilate never grows and fill-holes never absorbs, such as 1,4,8. In config files, use exclude_ids = [1, 4, 8].

--xy flag

Apply the operation to each z-section in 2D instead of in 3D.

--context / -c integer; default 64

The halo in voxels, in xy with --xy and otherwise in xy and z. The minimum is the iteration count for dilate and erode, twice the iteration count for opening and closing, and the width of the largest hole plus its wall for fill-holes.

--block-size / -b integer; default 2048

The write-block size in xy, in voxels, rounded to a multiple of the chunk size. In z, each block is one chunk deep.

--num-workers / -w integer; default 20

The number of daisy workers that process the blocks.

--in-array / -i path, required

The label array to read.

--out-array / -o path

The label array to write. Default: the input path with the suffix _outlier_filtered.

--num-std / -n float; default 3.0

Remove objects whose size differs from the mean by more than this many standard deviations.

--min-size integer; default 0

Compute the mean and standard deviation from the objects with at least this many voxels.

--num-workers / -w integer; default 20

The number of daisy workers that process the blocks.

--dry-run flag

Print the size cutoffs and the number of objects to remove without writing.

--in-array / -i path, required

The label array to read.

--out-array / -o path

The label array to write. Default: the input path with the suffix _remapped.

--remove-ids / -r text

Comma-separated label ids to remove (set to 0)

--merge-ids / -m text, repeatable

Comma-separated label ids to merge into the first id of the group. Repeat the option for more groups.

--num-workers / -w integer; default 20

The number of daisy workers that process the blocks.

--in-array / -i path, required

The label array to read.

--out-array / -o path

The label array to write. Default: the input path with the suffix _size_filtered.

--min-size integer; default 0

Remove objects smaller than this many voxels (dust)

--max-size integer

Remove objects larger than this many voxels. With 0 or no value, the command keeps large objects.

--num-workers / -w integer; default 20

The number of daisy workers that process the blocks.

--dry-run flag

Print size statistics and the number of objects to remove without writing.

--in-array / -i path, required

The label array to read.

--out-array / -o path

The label array to write. Default: the input path with the suffix _z_filtered.

--min-z / -z integer; default 1

Remove objects whose z-extent is this many slices or fewer

--num-workers / -w integer; default 20

The number of daisy workers that process the blocks.

--dry-run flag

Print the total object count and the number of objects to remove without writing.

--step text

The evaluate step to run when the file has several

--gt / -gt flag

Evaluate only against ground truth

--pred / -p flag

Evaluate only against predictions

--out-result / -o path

Result JSON path. Each volume gets its own file, and with several checks each file name also gets the check name.

--in-array / -i path, required

Path to the input Zarr array.

--out-array / -o path

Path to the output array

--padding / -p integer; default 0

Padding in voxels on each side of the bounding box, clipped to the input array.

--in-path / -i path, required

Path to the input 3D image, or to a directory of 2D images

--out-array / -o path, required

Path to the output Zarr array

--dtype / -d text; default uint8

Output data type

--voxel-size / -vs integer x 3; default 1 1 1

Size of each voxel in physical units (space-separated integers)

--voxel-offset / -vo integer x 3; default 0 0 0

Offset in voxels (space-separated integers)

--axis-names / -ax text x 3; default z y x

Names of the axes (space-separated strings)

--units / -u text x 3; default nm nm nm

Physical units of the axes (space-separated strings)

--crop / -c flag

Crop to the bounding box of positive voxels

--model-name / -m

The models to download, by name or short form, separated by commas. Example: -m 3Af2M

--setup-dir / -s

The directory to download the checkpoints to

--in-array / -i path, required

The path of the input Zarr array

--out-array / -o path

The path of the output mask Zarr array

--mode / -m raw or labels, required

raw masks an image, and labels masks objects

--in-seg / -i path, required

The path of the input segmentation Zarr array

--out-seg / -o path

The path of the output segmentation Zarr array

--luts / -l path, required

Path to the LUTs file

--in-array / -i path, required

The path of the input Zarr array, which may already end in a scale level

--scales / -s text, repeatable, required

Spatial scale factors for one level. Repeat -s for each level. Example: -s 2,2,2

--chunk-shape / -c text

Spatial chunk shape in voxels. The default is the input's. Example: -c 64,64,64

--mode / -m up or down, required

--labels / --image flag

--labels samples ids. --image averages intensities to downscale and interpolates them to upscale. The default is --labels for uint32, uint64, or a name with label, lbl, ids, mask or seg

--radius / -r integer; default 24

Disk radius in voxels: 24 at 4 nm (about 100 nm), 12 at 8 nm

--num-workers / -w integer; default 8

The number of worker processes. Allow about 80 bytes per voxel of one section (3.8 GB at 4926 x 9750).