Write a mask to a NumPy .npy file

Write an annot_mask as a lossless integer NumPy array, for interchange with Python tools. A <path>.legend.json sidecar (value -> label) is written alongside so the array stays self-describing. Unlike at_write_mask(), values are stored raw (no bit-depth normalisation), so bitfield artefact masks survive intact.

Usage

at_write_npy(
  mask,
  path,
  dtype = c("int32", "uint8", "int16"),
  transpose = FALSE,
  legend = TRUE,
  overwrite = FALSE,
  call = rlang::caller_env()
)

Arguments

mask

An annot_mask.

path

Output file path.

dtype

NumPy integer dtype: "int32" (default, wide enough for bitfield codes), "uint8", or "int16".

transpose

Logical; write the transpose, i.e. numpy shape (width, height) = [x, y], matching tools that store the native cube order (e.g. TIVITA/HyperGui). Default FALSE writes the standard image convention (height, width).

legend

Logical; also write a <path>.legend.json sidecar. Default TRUE.

overwrite

Logical; overwrite an existing file. Default FALSE.

call

The calling environment, for error reporting.

Value

The output path, invisibly.

See also

at_read_npy(), at_write_mask()

Other masks: at_mask(), at_mask_agreement(), at_mask_boundary(), at_mask_derive(), at_mask_legend(), at_mask_preview(), at_mask_stack(), at_mask_stats(), at_read_mask(), at_read_npy(), at_write_mask(), at_write_masks()

Examples

m <- at_mask(at_example_project(), "labelled")
f <- tempfile(fileext = ".npy")
at_write_npy(m, f)