77# ' - **Fashion-MNIST**: Clothing item images for classification.
88# ' - **Kuzushiji-MNIST**: Japanese cursive character dataset.
99# ' - **QMNIST**: Extended MNIST with high-precision NIST data.
10- # ' - **EMNIST**: Letters and digits with multiple label splits.
10+ # ' - **EMNIST**: A collection of letters and digits with multiple datasets and splits.
1111# '
1212# ' @param root Root directory for dataset storage. The dataset will be stored under `root/<dataset-name>`. Defaults to `tempdir()`.
1313# ' @param train Logical. If TRUE, use the training set; otherwise, use the test set. Not applicable to all datasets.
1818# '
1919# ' @return A torch dataset object, where each items is a list of `x` (image) and `y` (label).
2020# '
21- # ' @section Supported `kind `s for `emnist_dataset ()`:
21+ # ' @section Supported `dataset `s for `emnist_collection ()`:
2222# ' - `"byclass"`: 62 classes (digits + uppercase + lowercase)
2323# ' - `"bymerge"`: 47 classes (merged uppercase and lowercase)
2424# ' - `"balanced"`: 47 classes, balanced digits and letters
4343# ' item$x
4444# ' item$y
4545# '
46- # ' emnist <- emnist_dataset(kind = "balanced", split = "test", download = TRUE)
46+ # ' emnist <- emnist_collection(dataset = "balanced", split = "test", download = TRUE)
4747# ' item <- emnist[1]
4848# ' item$x
4949# ' item$y
@@ -354,19 +354,19 @@ fashion_mnist_dataset <- dataset(
354354 )
355355)
356356
357- # ' @describeIn mnist_dataset EMNIST dataset with digits and letters and multiple split modes .
358- # ' @param kind change the classes into one of "byclass", "bymerge", "balanced" representing the kind of emnist dataset. You
359- # ' can look at dataset attribute `$classes` to see the actual classes.
357+ # ' @describeIn mnist_dataset EMNIST collection with digits and letters arranged in multiple datasets .
358+ # ' @param dataset one of "byclass", "bymerge", "balanced" representing the subset of emnist collection
359+ # ' made of a set of classes. You can look at dataset attribute `$classes` to see the actual classes.
360360# ' @export
361- emnist_dataset <- dataset(
362- name = " emnist_dataset "
361+ emnist_collection <- dataset(
362+ name = " emnist_collection "
363363 archive_size = " 540 MB"
364364
365365 resources = list (
366366 c(" https://biometrics.nist.gov/cs_links/EMNIST/gzip.zip" " 58c8d27c78d21e728a6bc7b3cc06412e"
367367 ),
368368 rds_file = function (split , kind ) paste0(split ," -" kind ," .rds"
369- classes_all_kind = list (
369+ classes_all_dataset = list (
370370 byclass = c(
371371 " 0" " 1" " 2" " 3" " 4" " 5" " 6" " 7" " 8" " 9"
372372 LETTERS ,
@@ -390,34 +390,34 @@ emnist_dataset <- dataset(
390390 initialize = function (
391391 root = tempdir(),
392392 split = " test"
393- kind = " balanced"
393+ dataset = " balanced"
394394 transform = NULL ,
395395 target_transform = NULL ,
396396 download = FALSE
397397 ) {
398398
399399 self $ split <- match.arg(split , choices = c(" train" " test"
400- self $ kind <- match.arg(kind , choices = names(self $ classes_all_kind ))
400+ self $ dataset <- match.arg(dataset , choices = names(self $ classes_all_dataset ))
401401 self $ root_path <- root
402402 self $ raw_folder <- file.path(root , class(self )[1 ], " raw"
403403 self $ processed_folder <- file.path(root , class(self )[1 ], " processed"
404404 self $ transform <- transform
405405 self $ target_transform <- target_transform
406- self $ class <- self $ classes_all_kind [[self $ kind ]]
406+ self $ class <- self $ classes_all_dataset [[self $ dataset ]]
407407
408408 if (download ) {
409- cli_inform(" Dataset {.cls {class(self)[[1]]}} (~{.emph {self$archive_size}}) will be downloaded and processed if not already available."
409+ cli_inform(" {.cls {class(self)[[1]]}} (~{.emph {self$archive_size}}) will be downloaded and processed if not already available."
410410 self $ download()
411411 }
412412
413413 if (! self $ check_exists())
414414 runtime_error(" Dataset not found. You can use `download = TRUE` to download it."
415415
416- dataset_lst <- readRDS(file.path(self $ processed_folder , self $ rds_file(self $ split , self $ kind )))
416+ dataset_lst <- readRDS(file.path(self $ processed_folder , self $ rds_file(self $ split , self $ dataset )))
417417 self $ data <- dataset_lst [[1 ]]
418418 self $ targets <- dataset_lst [[2 ]] + 1L
419419
420- cli_inform(" Split {.val {self$split}} of {.cls {class( self)[[1]] }} dataset of kind {.val { self$kind }} processed successfully!"
420+ cli_inform(" Split {.val {self$split}} of dataset {.val { self$dataset }} from {.cls {class( self)[[1]] }} processed successfully!"
421421 },
422422
423423 download = function () {
@@ -440,15 +440,15 @@ emnist_dataset <- dataset(
440440 unzipped_root <- fs :: dir_ls(unzip_dir , type = " directory" recurse = FALSE )[1 ]
441441
442442 # only manage extraction of the 2 ubyte.gz under interest
443- img <- file.path(unzipped_root , glue :: glue(" emnist-{self$kind }-{self$split}-images-idx3-ubyte.gz"
444- lbl <- file.path(unzipped_root , glue :: glue(" emnist-{self$kind }-{self$split}-labels-idx1-ubyte.gz"
443+ img <- file.path(unzipped_root , glue :: glue(" emnist-{self$dataset }-{self$split}-images-idx3-ubyte.gz"
444+ lbl <- file.path(unzipped_root , glue :: glue(" emnist-{self$dataset }-{self$split}-labels-idx1-ubyte.gz"
445445 dataset_set <- list (read_sn3_pascalvincent(img ), read_sn3_pascalvincent(lbl ))
446- saveRDS(dataset_set , file.path(self $ processed_folder , self $ rds_file(self $ split , self $ kind )))
446+ saveRDS(dataset_set , file.path(self $ processed_folder , self $ rds_file(self $ split , self $ dataset )))
447447
448448 },
449449 # only manage existence of the rds file under interest
450450 check_exists = function () {
451- fs :: file_exists(file.path(self $ processed_folder , self $ rds_file(self $ split , self $ kind )))
451+ fs :: file_exists(file.path(self $ processed_folder , self $ rds_file(self $ split , self $ dataset )))
452452 },
453453
454454 .getitem = function (index ) {
@@ -492,3 +492,12 @@ read_sn3_pascalvincent <- function(path) {
492492 a <- aperm(a , perm = rev(seq_along(dim )))
493493 a
494494}
495+
496+ # ' @describeIn mnist_dataset Deprecated. Please use emnist_collection.
497+ # ' @param kind the `dataset` in `emnist_collection`.
498+ # ' @param ... the other `emnist_collection` parameters.
499+ # ' @export
500+ emnist_dataset <- function (kind , ... ){
501+ .Deprecated(" emnist_collection"
502+ emnist_collection(dataset = kind , ... )
503+ }
0 commit comments