library("rplatypus")
split <- split_dataset("scans", "splits", group_by = "^(patient\\\\d+)_")
validation <- split_path(split, "validation")
images <- read_images(validation$images, size = c(32, 32, 32), channels = 1)The files in one part of a split
Description
Reads the CSV that split_dataset() wrote and hands back the paths ready to use.
Usage
split_path(split, which = c("train", "validation", "test"))
Arguments
split
|
A split_dataset().
|
which
|
“train”, “validation” or “test”.
|
Details
The reason this exists rather than read.csv(): the CSVs store paths relative to themselves, so that a dataset and its split can be moved or mounted elsewhere together. Read directly, those paths do not open from wherever your session happens to be - which is a trap worth removing rather than documenting.
Value
A data frame with key, group, images and masks, the paths made absolute. images and masks hold one path per sample, or several separated by the split’s separator when a sample has several files.