Skip to contents

Estimate segmentation memory requirements

Source: R/estimate_segment_memory.R
estimate_segment_memory.Rd

Estimate the RAM pressure from the exact Ward backend before allocating the all-pairs distance vector. This is a lower bound: clustering and R object overhead require additional memory.

Usage

estimate_segment_memory(
  input,
  valid_mode = c("signal", "finite"),
  knn_k = 20,
  overhead_factor = 2
)

Arguments

input

A FITS-like object with imDat, a raw 3-D cube array, or an integer giving the number of valid pixels.

valid_mode

Valid-pixel rule used when input is a cube: "signal" matches segment, while "finite" requires every spectral channel to be finite.

knn_k

Number of nearest neighbours used for the sparse graph estimate.

overhead_factor

Multiplicative factor used for a conservative exact Ward RAM estimate beyond the condensed distance vector alone.

Value

A one-row data frame with valid-pixel count and approximate RAM estimates in GB.