init(labeledByNamesAt:separator:index:type:continueOnFailure:)
Reads training examples from a directory containing files having their labels in the name. The name can contain multiple words separated by a separator. So the index tells the position of the label in the file name. Files with incorrect name format are ignored.
Declaration
init(labeledByNamesAt url: URL, separator: Character = ".", index: Int = 0, type: UTType, continueOnFailure: Bool = false) throwsParameters
- url:
URL of directory containing the files.
- separator:
The separator used in the name. Default value is “.”
- index:
Index of the label in the file name. Default value is 0.
- type:
Type of files.
- continueOnFailure:
A Boolean value indicating whether to continue reading files after encountering a file that is not readable. The default value is
false.
Discussion
Take for example this directory structure:
/
fold1-foo-file1.png
fold1-foo-file2.png
fold2-foo-file3.png
fold1-bar-file4.png
fold1-bar-file5.png
fold2-bar-file6.pngWhen we specify separator as “-” and index as 1, it would produce two labels (foo and bar) with three URLs each.