diff --git a/DESCRIPTION b/DESCRIPTION
index 47fc9ec7..5fe92f4c 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -80,6 +80,7 @@ Collate:
     'dataset-places365.R'
     'dataset-plankton.R'
     'dataset-rf100-peixos.R'
+    'dataset-vggface2.R'
     'extension.R'
     'globals.R'
     'imagenet.R'
diff --git a/NAMESPACE b/NAMESPACE
index 90693b1b..7a027459 100644
--- a/NAMESPACE
+++ b/NAMESPACE
@@ -214,6 +214,7 @@ export(transform_rotate)
 export(transform_ten_crop)
 export(transform_to_tensor)
 export(transform_vflip)
+export(vggface2_dataset)
 export(vision_make_grid)
 export(whoi_plankton_dataset)
 export(whoi_small_coralnet_dataset)
diff --git a/NEWS.md b/NEWS.md
index ccd6202a..d199b4da 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -5,11 +5,13 @@
 * Added `lfw_people_dataset()` and `lfw_pairs_dataset()` for loading Labelled Faces in the Wild (LFW) datasets (@DerrickUnleashed, #203).
 * Added `places365_dataset()`for loading the Places365 dataset (@koshtiakanksha, #196).
 * Added `pascal_segmentation_dataset()`, and `pascal_detection_dataset()` for loading the Pascal Visual Object Classes datasets (@DerrickUnleashed, #209).
+* Added `whoi_plankton_dataset()`, and `whoi_small_plankton_dataset()` (@cregouby, #236).
 * Added `whoi_plankton_dataset()`, `whoi_small_plankton_dataset()`, and  `whoi_small_coral_dataset()` (@cregouby, #236).
 * Added `rf100_document_collection()`, `rf100_medical_collection()`, `rf100_biology_collection()`, `rf100_damage_collection()`, `rf100_infrared_collection()`, 
   and `rf100_underwater_collection()` . Those are collection of datasets from RoboFlow 100 under the same 
   thematic, for a total of 35 datasets (@koshtiakanksha, @cregouby, #239).
 * Added `rf100_peixos_segmentation_dataset()`.  (@koshtiakanksha, @cregouby, #250).
+* Added `vggface2_dataset()` for loading the VGGFace2 dataset (@DerrickUnleashed, #238).
 
 ## New models
 
diff --git a/R/dataset-vggface2.R b/R/dataset-vggface2.R
new file mode 100644
index 00000000..7019989b
--- /dev/null
+++ b/R/dataset-vggface2.R
@@ -0,0 +1,185 @@
+#' VGGFace2 Dataset
+#'
+#' The VGGFace2 dataset is a large-scale face recognition dataset containing images
+#' of celebrities from a wide range of ethnicities, professions, and ages.
+#' Each identity has multiple images with large variations in pose, age, illumination,
+#' ethnicity, and profession.
+#'
+#' @inheritParams oxfordiiitpet_dataset
+#' @param root Character. Root directory where the dataset will be stored under `root/vggface2`.
+#'
+#' @return A torch dataset object `vggface2_dataset`:
+#' - `x`: RGB image array.
+#' - `y`: Integer label (1…N) for the identity.
+#' 
+#' `ds$classes` is a named list mapping integer labels to a list with:
+#' - `name`: Character name of the person.
+#' - `gender`: "Male" or "Female".
+#'
+#' @examples
+#' \dontrun{
+#' #Load the training set
+#' ds <- vggface2_dataset(download = TRUE)
+#' item <- ds[1]
+#' item$x      # image array RGB
+#' item$y      # integer label
+#' ds$classes[item$y]  # list(name=..., gender=...)
+#'
+#' #Load the test set
+#' ds <- vggface2_dataset(download = TRUE, train = FALSE)
+#' item <- ds[1]
+#' item$x      # image array RGB
+#' item$y      # integer label
+#' ds$classes[item$y]  # list(name=..., gender=...)
+#' }
+#'
+#' @family segmentation_dataset
+#' @export
+vggface2_dataset <- torch::dataset(
+  name = "vggface2",
+  resources = data.frame(
+    split = c("train_images", "test_images", "train_list", "test_list", "identity"),
+    url = c(
+      "https://huggingface.co/datasets/ProgramComputer/VGGFace2/resolve/main/data/vggface2_train.tar.gz",
+      "https://huggingface.co/datasets/ProgramComputer/VGGFace2/resolve/main/data/vggface2_test.tar.gz",
+      "https://huggingface.co/datasets/ProgramComputer/VGGFace2/resolve/main/meta/train_list.txt",
+      "https://huggingface.co/datasets/ProgramComputer/VGGFace2/resolve/main/meta/test_list.txt",
+      "https://huggingface.co/datasets/ProgramComputer/VGGFace2/raw/main/meta/identity_meta.csv"
+    ),
+    md5 = c(
+      "88813c6b15de58afc8fa75ea83361d7f",
+      "bb7a323824d1004e14e00c23974facd3",
+      "4cfbab4a839163f454d7ecef28b68669",
+      "d08b10f12bc9889509364ef56d73c621",
+      "d315386c7e8e166c4f60e27d9cc61acc"
+    )
+  ),
+  training_file = "train.rds",
+  test_file = "test.rds",
+
+  initialize = function(
+    root = tempdir(),
+    train = TRUE,
+    transform = NULL,
+    target_transform = NULL,
+    download = FALSE
+  ) {
+    self$root_path <- root
+    self$transform <- transform
+    self$target_transform <- target_transform
+    if (train) {
+      self$split <- "train"
+      self$archive_size <- "36 GB"
+    } else {
+      self$split <- "test"
+      self$archive_size <- "2 GB"
+    }
+
+    if (download) {
+      cli_inform("Dataset {.cls {class(self)[[1]]}} (~{.emph {self$archive_size}}) will be downloaded and processed if not already available.")
+      self$download()
+    }
+
+    if (!self$check_exists()) {
+      cli_abort("Dataset not found. You can use `download = TRUE` to download it.")
+    }
+
+    if (train) {
+      data_file <- self$training_file
+    } else {
+      data_file <- self$test_file
+    }
+    data <- readRDS(file.path(self$processed_folder, data_file))
+
+    self$img_path <- data$img_path
+    self$labels <- data$labels
+    self$classes <- data$classes
+
+    cli_inform("{.cls {class(self)[[1]]}} dataset loaded with {self$.length()} images across {length(self$classes)} classes.")
+  },
+
+  download = function() {
+    if (self$check_exists()) {
+      return()
+    }
+
+    fs::dir_create(self$raw_folder)
+    fs::dir_create(self$processed_folder)
+
+    cli_inform("Downloading {.cls {class(self)[[1]]}}...")
+
+    for (i in seq_len(nrow(self$resources))) {
+      row <- self$resources[i, ]
+      archive <- download_and_cache(row$url, prefix = row$split)
+      if (tools::md5sum(archive) != row$md5) {
+        runtime_error("Corrupt file! Delete the file in {archive} and try again.")
+      }
+      if (tools::file_ext(row$url) == "gz") {
+        utils::untar(archive, exdir = self$raw_folder)
+      } else {
+        fs::file_move(archive, self$raw_folder)
+      }
+    }
+    
+    identity_file <- file.path(self$raw_folder, "identity_meta.csv")
+    identity_df <- read.csv(identity_file, sep = ",", stringsAsFactors = FALSE, strip.white = TRUE)
+    identity_df$Class_ID <- trimws(identity_df$Class_ID)
+    identity_df$Gender <- factor(identity_df$Gender, levels = c("f", "m"), labels = c("Female", "Male"))
+
+    for (split in c("train", "test")) {
+      if (split == "train") {
+        list_file <- file.path(self$raw_folder, "train_list.txt")
+      } else {
+        list_file <- file.path(self$raw_folder, "test_list.txt")
+      }
+
+      split_df <- read.delim(
+        list_file,
+        sep = "/",
+        col.names = c("Class_ID", "img_path"),
+        header = FALSE,
+        stringsAsFactors = FALSE
+      )
+
+      merged_df <- merge(split_df, identity_df, by = "Class_ID", all.x = TRUE)
+      merged_df$Label <- as.integer(factor(merged_df$Class_ID, levels = unique(merged_df$Class_ID)))
+
+      saveRDS(
+        merged_df,
+        file.path(self$processed_folder, paste0(split, ".rds"))
+      )
+    }
+    cli_inform("Dataset {.cls {class(self)[[1]]}} downloaded and extracted successfully.")
+  },
+
+  check_exists = function() {
+    fs::file_exists(file.path(self$processed_folder, self$training_file)) &&
+      fs::file_exists(file.path(self$processed_folder, self$test_file))
+  },
+
+  .getitem = function(index) {
+    x <- jpeg::readJPEG(self$img_path[index])
+    y <- self$labels[index]
+
+    if (!is.null(self$transform)) {
+      x <- self$transform(x)
+    }
+    if (!is.null(self$target_transform)) {
+      y <- self$target_transform(y)
+    }
+    list(x = x, y = y)
+  },
+
+  .length = function() {
+    length(self$img_path)
+  },
+
+  active = list(
+    raw_folder = function() {
+      file.path(self$root_path, "vggface2", "raw")
+    },
+    processed_folder = function() {
+      file.path(self$root_path, "vggface2", "processed")
+    }
+  )
+)
\ No newline at end of file
diff --git a/man/oxfordiiitpet_segmentation_dataset.Rd b/man/oxfordiiitpet_segmentation_dataset.Rd
index c1a059ad..aae686e2 100644
--- a/man/oxfordiiitpet_segmentation_dataset.Rd
+++ b/man/oxfordiiitpet_segmentation_dataset.Rd
@@ -69,5 +69,6 @@ tensor_image_browse(overlay)
 Other segmentation_dataset: 
 \code{\link{pascal_voc_datasets}},
 \code{\link{rf100_peixos_segmentation_dataset}()}
+\code{\link{vggface2_dataset}()}
 }
 \concept{segmentation_dataset}
diff --git a/man/pascal_voc_datasets.Rd b/man/pascal_voc_datasets.Rd
index 24c198e4..f5d0dd74 100644
--- a/man/pascal_voc_datasets.Rd
+++ b/man/pascal_voc_datasets.Rd
@@ -130,6 +130,7 @@ tensor_image_browse(boxed_img)
 Other segmentation_dataset: 
 \code{\link{oxfordiiitpet_segmentation_dataset}()},
 \code{\link{rf100_peixos_segmentation_dataset}()}
+\code{\link{vggface2_dataset}()}
 
 Other detection_dataset: 
 \code{\link{coco_detection_dataset}()},
diff --git a/man/vggface2_dataset.Rd b/man/vggface2_dataset.Rd
new file mode 100644
index 00000000..db558d87
--- /dev/null
+++ b/man/vggface2_dataset.Rd
@@ -0,0 +1,68 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/dataset-vggface2.R
+\name{vggface2_dataset}
+\alias{vggface2_dataset}
+\title{VGGFace2 Dataset}
+\usage{
+vggface2_dataset(
+  root = tempdir(),
+  train = TRUE,
+  transform = NULL,
+  target_transform = NULL,
+  download = FALSE
+)
+}
+\arguments{
+\item{root}{Character. Root directory where the dataset will be stored under \code{root/vggface2}.}
+
+\item{train}{Logical. If TRUE, use the training set; otherwise, use the test set. Not applicable to all datasets.}
+
+\item{transform}{Optional. A function that takes an image and returns a transformed version (e.g., normalization, cropping).}
+
+\item{target_transform}{Optional. A function that transforms the label.}
+
+\item{download}{Logical. If TRUE, downloads the dataset to \verb{root/}. If the dataset is already present, download is skipped.}
+}
+\value{
+A torch dataset object \code{vggface2_dataset}:
+\itemize{
+\item \code{x}: RGB image array.
+\item \code{y}: Integer label (1…N) for the identity.
+}
+
+\code{ds$classes} is a named list mapping integer labels to a list with:
+\itemize{
+\item \code{name}: Character name of the person.
+\item \code{gender}: "Male" or "Female".
+}
+}
+\description{
+The VGGFace2 dataset is a large-scale face recognition dataset containing images
+of celebrities from a wide range of ethnicities, professions, and ages.
+Each identity has multiple images with large variations in pose, age, illumination,
+ethnicity, and profession.
+}
+\examples{
+\dontrun{
+#Load the training set
+ds <- vggface2_dataset(download = TRUE)
+item <- ds[1]
+item$x      # image array RGB
+item$y      # integer label
+ds$classes[item$y]  # list(name=..., gender=...)
+
+#Load the test set
+ds <- vggface2_dataset(download = TRUE, train = FALSE)
+item <- ds[1]
+item$x      # image array RGB
+item$y      # integer label
+ds$classes[item$y]  # list(name=..., gender=...)
+}
+
+}
+\seealso{
+Other segmentation_dataset: 
+\code{\link{oxfordiiitpet_segmentation_dataset}()},
+\code{\link{pascal_voc_datasets}}
+}
+\concept{segmentation_dataset}
diff --git a/tests/testthat/test-dataset-vggface2.R b/tests/testthat/test-dataset-vggface2.R
new file mode 100644
index 00000000..b4b57ee3
--- /dev/null
+++ b/tests/testthat/test-dataset-vggface2.R
@@ -0,0 +1,32 @@
+context('dataset-vggface2')
+
+t <- withr::local_tempdir()
+options(timeout = 60000)
+
+test_that("VGGFace2 dataset works correctly for train split", {
+
+  skip_if(Sys.getenv("TEST_LARGE_DATASETS", unset = 0) != 1,
+          "Skipping test: set TEST_LARGE_DATASETS=1 to enable tests requiring large downloads.")
+
+  vgg <- vggface2_dataset(root = t, download = TRUE)
+  expect_length(vgg, 3141890)
+  first_item <- vgg[1]
+  expect_named(first_item, c("x", "y"))
+  expect_type(first_item$x, "double")
+  expect_type(first_item$y, "integer")
+  expect_equal(first_item$y, 1)
+})
+
+test_that("VGGFace2 dataset works correctly for test split", {
+
+  skip_if(Sys.getenv("TEST_LARGE_DATASETS", unset = 0) != 1,
+          "Skipping test: set TEST_LARGE_DATASETS=1 to enable tests requiring large downloads.")
+
+  vgg <- vggface2_dataset(root = t, train = FALSE)
+  expect_length(vgg, 169396)
+  first_item <- vgg[1]
+  expect_named(first_item, c("x", "y"))
+  expect_type(first_item$x, "double")
+  expect_type(first_item$y, "integer")
+  expect_equal(first_item$y, 1)
+})
\ No newline at end of file