diff --git a/setup.py b/setup.py
index 09f951b1090..53bba409407 100644
--- a/setup.py
+++ b/setup.py
@@ -495,7 +495,7 @@ def run(self):
         # Package info
         packages=find_packages(exclude=('test',)),
         package_data={
-            package_name: ['*.dll', '*.dylib', '*.so']
+            package_name: ['*.dll', '*.dylib', '*.so', 'prototype/datasets/_builtin/*.categories']
         },
         zip_safe=False,
         install_requires=requirements,
diff --git a/test/prototype/datasets/conftest.py b/test/prototype/datasets/conftest.py
new file mode 100644
index 00000000000..474d5fb6240
--- /dev/null
+++ b/test/prototype/datasets/conftest.py
@@ -0,0 +1,10 @@
+import pytest
+from torchvision.prototype.datasets.utils import DatasetInfo
+
+
+@pytest.fixture
+def make_minimal_dataset_info():
+    def make(name="name", categories=None, **kwargs):
+        return DatasetInfo(name, categories=categories or [], **kwargs)
+
+    return make
diff --git a/test/prototype/datasets/test_api.py b/test/prototype/datasets/test_api.py
index 5ba432fd1f4..4e2102ee320 100644
--- a/test/prototype/datasets/test_api.py
+++ b/test/prototype/datasets/test_api.py
@@ -3,7 +3,7 @@
 from torchvision.prototype import datasets
 from torchvision.prototype.datasets import _api
 from torchvision.prototype.datasets import _builtin
-from torchvision.prototype.datasets.utils import Dataset, DatasetInfo, DatasetConfig
+from torchvision.prototype.datasets.utils import Dataset, DatasetConfig
 
 
 @pytest.fixture
@@ -14,9 +14,9 @@ def patch_datasets(monkeypatch):
 
 
 @pytest.fixture
-def dataset(mocker):
-    info = DatasetInfo(
-        "name", valid_options=dict(split=("train", "test"), foo=("bar", "baz"))
+def dataset(mocker, make_minimal_dataset_info):
+    info = make_minimal_dataset_info(
+        valid_options=dict(split=("train", "test"), foo=("bar", "baz"))
     )
 
     class DatasetMock(Dataset):
diff --git a/test/prototype/datasets/test_utils.py b/test/prototype/datasets/test_utils.py
index 4cac156bc43..14673efef03 100644
--- a/test/prototype/datasets/test_utils.py
+++ b/test/prototype/datasets/test_utils.py
@@ -122,42 +122,43 @@ class TestDatasetInfo:
     def valid_options():
         return dict(split=("train", "test"), foo=("bar", "baz"))
 
-    def test_no_valid_options(self):
-        info = utils.DatasetInfo("name")
+    @staticmethod
+    @pytest.fixture
+    def info(make_minimal_dataset_info, valid_options):
+        return make_minimal_dataset_info(valid_options=valid_options)
+
+    def test_no_valid_options(self, make_minimal_dataset_info):
+        info = make_minimal_dataset_info()
         assert info.default_config.split == "train"
 
-    def test_valid_options_no_split(self):
-        info = utils.DatasetInfo("name", valid_options=dict(option=("argument",)))
+    def test_valid_options_no_split(self, make_minimal_dataset_info):
+        info = make_minimal_dataset_info(valid_options=dict(option=("argument",)))
         assert info.default_config.split == "train"
 
-    def test_valid_options_no_train(self):
+    def test_valid_options_no_train(self, make_minimal_dataset_info):
         with pytest.raises(ValueError):
-            utils.DatasetInfo("name", valid_options=dict(split=("test",)))
+            make_minimal_dataset_info(valid_options=dict(split=("test",)))
 
-    def test_default_config(self, valid_options):
+    def test_default_config(self, make_minimal_dataset_info, valid_options):
         default_config = utils.DatasetConfig(
             {key: values[0] for key, values in valid_options.items()}
         )
 
         assert (
-            utils.DatasetInfo("name", valid_options=valid_options).default_config
+            make_minimal_dataset_info(valid_options=valid_options).default_config
             == default_config
         )
 
-    def test_make_config_unknown_option(self, valid_options):
-        info = utils.DatasetInfo("name", valid_options=valid_options)
-
+    def test_make_config_unknown_option(self, info):
         with pytest.raises(ValueError):
             info.make_config(unknown_option=None)
 
-    def test_make_config_invalid_argument(self, valid_options):
-        info = utils.DatasetInfo("name", valid_options=valid_options)
-
+    def test_make_config_invalid_argument(self, info):
         with pytest.raises(ValueError):
             info.make_config(split="unknown_split")
 
-    def test_repr(self, valid_options):
-        output = repr(utils.DatasetInfo("name", valid_options=valid_options))
+    def test_repr(self, make_minimal_dataset_info, valid_options):
+        output = repr(make_minimal_dataset_info(valid_options=valid_options))
 
         assert isinstance(output, str)
         assert "DatasetInfo" in output
@@ -165,9 +166,9 @@ def test_repr(self, valid_options):
             assert f"{key}={value}" in output
 
     @pytest.mark.parametrize("optional_info", ("citation", "homepage", "license"))
-    def test_repr_optional_info(self, optional_info):
+    def test_repr_optional_info(self, make_minimal_dataset_info, optional_info):
         sentinel = "sentinel"
-        info = utils.DatasetInfo("name", **{optional_info: sentinel})
+        info = make_minimal_dataset_info(**{optional_info: sentinel})
 
         assert f"{optional_info}={sentinel}" in repr(info)
 
@@ -183,6 +184,7 @@ def make(name="name", valid_options=None, resources=None):
                 dict(
                     info=utils.DatasetInfo(
                         name,
+                        categories=[],
                         valid_options=valid_options or dict(split=("train", "test")),
                     ),
                     resources=mocker.Mock(return_value=[])
diff --git a/torchvision/prototype/datasets/__init__.py b/torchvision/prototype/datasets/__init__.py
index 7bd35c72594..f35a3776d05 100644
--- a/torchvision/prototype/datasets/__init__.py
+++ b/torchvision/prototype/datasets/__init__.py
@@ -1,3 +1,4 @@
+from ._home import *
 from . import decoder, utils, datapipes
 
 # Load this last, since itself but especially _builtin/* depends on the above being available
diff --git a/torchvision/prototype/datasets/_api.py b/torchvision/prototype/datasets/_api.py
index 46c21a596d2..025b0927168 100644
--- a/torchvision/prototype/datasets/_api.py
+++ b/torchvision/prototype/datasets/_api.py
@@ -1,36 +1,17 @@
-import difflib
 import io
-import os
-import pathlib
-from typing import Any, Callable, Dict, List, Optional, Union
+from typing import Any, Callable, Dict, List, Optional
 
 import torch
-from torch.hub import _get_torch_home
 from torch.utils.data import IterDataPipe
 
+from torchvision.prototype.datasets import home
 from torchvision.prototype.datasets.decoder import pil
 from torchvision.prototype.datasets.utils import Dataset, DatasetInfo
 from torchvision.prototype.datasets.utils._internal import add_suggestion
 from . import _builtin
 
-__all__ = ["home", "register", "list", "info", "load"]
 
-
-# TODO: This needs a better default
-HOME = pathlib.Path(_get_torch_home()) / "datasets" / "vision"
-
-
-def home(home: Optional[Union[str, pathlib.Path]] = None) -> pathlib.Path:
-    global HOME
-    if home is not None:
-        HOME = pathlib.Path(home).expanduser().resolve()
-        return HOME
-
-    home = os.getenv("TORCHVISION_DATASETS_HOME")
-    if home is not None:
-        return pathlib.Path(home)
-
-    return HOME
+__all__ = ["register", "list", "info", "load"]
 
 
 DATASETS: Dict[str, Dataset] = {}
diff --git a/torchvision/prototype/datasets/_builtin/caltech256.categories b/torchvision/prototype/datasets/_builtin/caltech256.categories
new file mode 100644
index 00000000000..82128efba97
--- /dev/null
+++ b/torchvision/prototype/datasets/_builtin/caltech256.categories
@@ -0,0 +1,257 @@
+ak47
+american-flag
+backpack
+baseball-bat
+baseball-glove
+basketball-hoop
+bat
+bathtub
+bear
+beer-mug
+billiards
+binoculars
+birdbath
+blimp
+bonsai-101
+boom-box
+bowling-ball
+bowling-pin
+boxing-glove
+brain-101
+breadmaker
+buddha-101
+bulldozer
+butterfly
+cactus
+cake
+calculator
+camel
+cannon
+canoe
+car-tire
+cartman
+cd
+centipede
+cereal-box
+chandelier-101
+chess-board
+chimp
+chopsticks
+cockroach
+coffee-mug
+coffin
+coin
+comet
+computer-keyboard
+computer-monitor
+computer-mouse
+conch
+cormorant
+covered-wagon
+cowboy-hat
+crab-101
+desk-globe
+diamond-ring
+dice
+dog
+dolphin-101
+doorknob
+drinking-straw
+duck
+dumb-bell
+eiffel-tower
+electric-guitar-101
+elephant-101
+elk
+ewer-101
+eyeglasses
+fern
+fighter-jet
+fire-extinguisher
+fire-hydrant
+fire-truck
+fireworks
+flashlight
+floppy-disk
+football-helmet
+french-horn
+fried-egg
+frisbee
+frog
+frying-pan
+galaxy
+gas-pump
+giraffe
+goat
+golden-gate-bridge
+goldfish
+golf-ball
+goose
+gorilla
+grand-piano-101
+grapes
+grasshopper
+guitar-pick
+hamburger
+hammock
+harmonica
+harp
+harpsichord
+hawksbill-101
+head-phones
+helicopter-101
+hibiscus
+homer-simpson
+horse
+horseshoe-crab
+hot-air-balloon
+hot-dog
+hot-tub
+hourglass
+house-fly
+human-skeleton
+hummingbird
+ibis-101
+ice-cream-cone
+iguana
+ipod
+iris
+jesus-christ
+joy-stick
+kangaroo-101
+kayak
+ketch-101
+killer-whale
+knife
+ladder
+laptop-101
+lathe
+leopards-101
+license-plate
+lightbulb
+light-house
+lightning
+llama-101
+mailbox
+mandolin
+mars
+mattress
+megaphone
+menorah-101
+microscope
+microwave
+minaret
+minotaur
+motorbikes-101
+mountain-bike
+mushroom
+mussels
+necktie
+octopus
+ostrich
+owl
+palm-pilot
+palm-tree
+paperclip
+paper-shredder
+pci-card
+penguin
+people
+pez-dispenser
+photocopier
+picnic-table
+playing-card
+porcupine
+pram
+praying-mantis
+pyramid
+raccoon
+radio-telescope
+rainbow
+refrigerator
+revolver-101
+rifle
+rotary-phone
+roulette-wheel
+saddle
+saturn
+school-bus
+scorpion-101
+screwdriver
+segway
+self-propelled-lawn-mower
+sextant
+sheet-music
+skateboard
+skunk
+skyscraper
+smokestack
+snail
+snake
+sneaker
+snowmobile
+soccer-ball
+socks
+soda-can
+spaghetti
+speed-boat
+spider
+spoon
+stained-glass
+starfish-101
+steering-wheel
+stirrups
+sunflower-101
+superman
+sushi
+swan
+swiss-army-knife
+sword
+syringe
+tambourine
+teapot
+teddy-bear
+teepee
+telephone-box
+tennis-ball
+tennis-court
+tennis-racket
+theodolite
+toaster
+tomato
+tombstone
+top-hat
+touring-bike
+tower-pisa
+traffic-light
+treadmill
+triceratops
+tricycle
+trilobite-101
+tripod
+t-shirt
+tuning-fork
+tweezer
+umbrella-101
+unicorn
+vcr
+video-projector
+washing-machine
+watch-101
+waterfall
+watermelon
+welding-mask
+wheelbarrow
+windmill
+wine-bottle
+xylophone
+yarmulke
+yo-yo
+zebra
+airplanes-101
+car-side-101
+faces-easy-101
+greyhound
+tennis-shoes
+toad
+clutter
diff --git a/torchvision/prototype/datasets/_builtin/caltech256.py b/torchvision/prototype/datasets/_builtin/caltech256.py
index b1b2c87428f..57a32a5e1b2 100644
--- a/torchvision/prototype/datasets/_builtin/caltech256.py
+++ b/torchvision/prototype/datasets/_builtin/caltech256.py
@@ -7,7 +7,6 @@
 from torch.utils.data.datapipes.iter import (
     Mapper,
     TarArchiveReader,
-    FileLoader,
 )
 
 from torchvision.prototype.datasets.utils import (
@@ -16,6 +15,11 @@
     DatasetInfo,
     HttpResource,
 )
+from torchvision.prototype.datasets.utils._internal import create_categories_file
+
+__all__ = ["Caltech256"]
+
+HERE = pathlib.Path(__file__).parent
 
 
 class Caltech256(Dataset):
@@ -23,6 +27,7 @@ class Caltech256(Dataset):
     def info(self) -> DatasetInfo:
         return DatasetInfo(
             "caltech256",
+            categories=HERE / "caltech256.categories",
             homepage="http://www.vision.caltech.edu/Image_Datasets/Caltech256",
         )
 
@@ -63,3 +68,18 @@ def _make_datapipe(
         return Mapper(
             dp, self._collate_and_decode_sample, fn_kwargs=dict(decoder=decoder)
         )
+
+    def generate_categories_file(self, root):
+        dp = self.resources(self.default_config)[0].to_datapipe(
+            pathlib.Path(root) / self.name
+        )
+        dp = TarArchiveReader(dp)
+        dir_names = {pathlib.Path(path).parent.name for path, _ in dp}
+        categories = [name.split(".")[1] for name in sorted(dir_names)]
+        create_categories_file(HERE, self.name, categories)
+
+
+if __name__ == "__main__":
+    from torchvision.prototype.datasets import home
+
+    Caltech256().generate_categories_file(home())
diff --git a/torchvision/prototype/datasets/_builtin/cifar.py b/torchvision/prototype/datasets/_builtin/cifar.py
index 028cd5d5738..d8a88c142bb 100644
--- a/torchvision/prototype/datasets/_builtin/cifar.py
+++ b/torchvision/prototype/datasets/_builtin/cifar.py
@@ -2,8 +2,9 @@
 import functools
 import io
 import os.path
+import pathlib
 import pickle
-from typing import Any, Callable, Dict, List, Optional, Tuple
+from typing import Any, Callable, Dict, List, Optional, Tuple, Union
 
 import numpy as np
 import PIL.Image
@@ -24,6 +25,11 @@
     DatasetInfo,
     HttpResource,
 )
+from torchvision.prototype.datasets.utils._internal import create_categories_file
+
+__all__ = ["Cifar10", "Cifar100"]
+
+HERE = pathlib.Path(__file__).parent
 
 
 class _CifarBase(Dataset):
@@ -96,12 +102,25 @@ def _make_datapipe(
         dp = Zipper(images_dp, labels_dp)
         return Mapper(dp, self._collate_and_decode, fn_kwargs=dict(decoder=decoder))
 
+    def generate_categories_file(
+        self, root: Union[str, pathlib.Path], *, file_name: str, categories_key: str
+    ) -> None:
+        dp = self.resources(self.default_config)[0].to_datapipe(
+            pathlib.Path(root) / self.name
+        )
+        dp = TarArchiveReader(dp)
+        dp = Filter(dp, lambda data: os.path.basename(data[0]) == file_name)
+        dp = Mapper(dp, self._unpickle)
+        categories = next(iter(dp))[categories_key]
+        create_categories_file(HERE, self.name, categories)
+
 
 class Cifar10(_CifarBase):
     @property
     def info(self) -> DatasetInfo:
         return DatasetInfo(
             "cifar10",
+            categories=HERE / "cifar10.categories",
             homepage="https://www.cs.toronto.edu/~kriz/cifar.html",
         )
 
@@ -129,12 +148,24 @@ def _split_data_file(self, data: Tuple[str, Any]) -> Optional[int]:
         else:
             return None
 
+    def generate_categories_file(
+        self,
+        root: Union[str, pathlib.Path],
+        *,
+        file_name: str = "batches.meta",
+        categories_key: str = "label_names",
+    ) -> None:
+        super().generate_categories_file(
+            root, file_name=file_name, categories_key=categories_key
+        )
+
 
 class Cifar100(_CifarBase):
     @property
     def info(self) -> DatasetInfo:
         return DatasetInfo(
             "cifar100",
+            categories=HERE / "cifar100.categories",
             homepage="https://www.cs.toronto.edu/~kriz/cifar.html",
             valid_options=dict(
                 split=("train", "test"),
@@ -164,3 +195,22 @@ def _split_data_file(self, data: Tuple[str, Any]) -> Optional[int]:
             return 1
         else:
             return None
+
+    def generate_categories_file(
+        self,
+        root: Union[str, pathlib.Path],
+        *,
+        file_name: str = "meta",
+        categories_key: str = "fine_label_names",
+    ) -> None:
+        super().generate_categories_file(
+            root, file_name=file_name, categories_key=categories_key
+        )
+
+
+if __name__ == "__main__":
+    from torchvision.prototype.datasets import home
+
+    root = home()
+    for cls in (Cifar10, Cifar100):
+        cls().generate_categories_file(root)
diff --git a/torchvision/prototype/datasets/_builtin/cifar10.categories b/torchvision/prototype/datasets/_builtin/cifar10.categories
new file mode 100644
index 00000000000..fa30c22b95d
--- /dev/null
+++ b/torchvision/prototype/datasets/_builtin/cifar10.categories
@@ -0,0 +1,10 @@
+airplane
+automobile
+bird
+cat
+deer
+dog
+frog
+horse
+ship
+truck
diff --git a/torchvision/prototype/datasets/_builtin/cifar100.categories b/torchvision/prototype/datasets/_builtin/cifar100.categories
new file mode 100644
index 00000000000..7f7bf51d1ab
--- /dev/null
+++ b/torchvision/prototype/datasets/_builtin/cifar100.categories
@@ -0,0 +1,100 @@
+apple
+aquarium_fish
+baby
+bear
+beaver
+bed
+bee
+beetle
+bicycle
+bottle
+bowl
+boy
+bridge
+bus
+butterfly
+camel
+can
+castle
+caterpillar
+cattle
+chair
+chimpanzee
+clock
+cloud
+cockroach
+couch
+crab
+crocodile
+cup
+dinosaur
+dolphin
+elephant
+flatfish
+forest
+fox
+girl
+hamster
+house
+kangaroo
+keyboard
+lamp
+lawn_mower
+leopard
+lion
+lizard
+lobster
+man
+maple_tree
+motorcycle
+mountain
+mouse
+mushroom
+oak_tree
+orange
+orchid
+otter
+palm_tree
+pear
+pickup_truck
+pine_tree
+plain
+plate
+poppy
+porcupine
+possum
+rabbit
+raccoon
+ray
+road
+rocket
+rose
+sea
+seal
+shark
+shrew
+skunk
+skyscraper
+snail
+snake
+spider
+squirrel
+streetcar
+sunflower
+sweet_pepper
+table
+tank
+telephone
+television
+tiger
+tractor
+train
+trout
+tulip
+turtle
+wardrobe
+whale
+willow_tree
+wolf
+woman
+worm
diff --git a/torchvision/prototype/datasets/_builtin/mnist.py b/torchvision/prototype/datasets/_builtin/mnist.py
index 426da3b03ae..68bd7f82829 100644
--- a/torchvision/prototype/datasets/_builtin/mnist.py
+++ b/torchvision/prototype/datasets/_builtin/mnist.py
@@ -23,8 +23,10 @@
     DatasetInfo,
     HttpResource,
     OnlineResource,
-    image_buffer_from_array,
 )
+from torchvision.prototype.datasets.utils._internal import image_buffer_from_array
+
+__all__ = ["MNIST", "FashionMNIST", "KMNIST", "EMNIST", "QMNIST"]
 
 prod = functools.partial(functools.reduce, operator.mul)
 
@@ -132,6 +134,7 @@ class MNIST(_MNISTBase):
     def info(self):
         return DatasetInfo(
             "mnist",
+            categories=10,
             homepage="http://yann.lecun.com/exdb/mnist",
             valid_options=dict(
                 split=("train", "test"),
@@ -163,6 +166,18 @@ class FashionMNIST(MNIST):
     def info(self):
         return DatasetInfo(
             "fashionmnist",
+            categories=[
+                "T-shirt/top",
+                "Trouser",
+                "Pullover",
+                "Dress",
+                "Coat",
+                "Sandal",
+                "Shirt",
+                "Sneaker",
+                "Bag",
+                "Ankle boot",
+            ],
             homepage="https://github.com/zalandoresearch/fashion-mnist",
             valid_options=dict(
                 split=("train", "test"),
@@ -183,6 +198,7 @@ class KMNIST(MNIST):
     def info(self):
         return DatasetInfo(
             "kmnist",
+            categories=["o", "ki", "su", "tsu", "na", "ha", "ma", "ya", "re", "wo"],
             homepage="http://codh.rois.ac.jp/kmnist/index.html.en",
             valid_options=dict(
                 split=("train", "test"),
@@ -198,11 +214,16 @@ def info(self):
     }
 
 
+import string
+
+
 class EMNIST(_MNISTBase):
     @property
     def info(self):
         return DatasetInfo(
             "emnist",
+            # FIXME: shift the labels at runtime to always produce this
+            categories=list(string.digits + string.ascii_letters),
             homepage="https://www.westernsydney.edu.au/icns/reproducible_research/publication_support_materials/emnist",
             valid_options=dict(
                 split=("train", "test"),
@@ -276,6 +297,7 @@ class QMNIST(_MNISTBase):
     def info(self):
         return DatasetInfo(
             "qmnist",
+            categories=10,
             homepage="https://github.com/facebookresearch/qmnist",
             valid_options=dict(
                 split=("train", "test", "test10k", "test50k", "nist"),
diff --git a/torchvision/prototype/datasets/_home.py b/torchvision/prototype/datasets/_home.py
new file mode 100644
index 00000000000..333d570b9e1
--- /dev/null
+++ b/torchvision/prototype/datasets/_home.py
@@ -0,0 +1,21 @@
+import os
+import pathlib
+from typing import Optional, Union
+
+from torch.hub import _get_torch_home
+
+# TODO: This needs a better default
+HOME = pathlib.Path(_get_torch_home()) / "datasets" / "vision"
+
+
+def home(home: Optional[Union[str, pathlib.Path]] = None) -> pathlib.Path:
+    global HOME
+    if home is not None:
+        HOME = pathlib.Path(home).expanduser().resolve()
+        return HOME
+
+    home = os.getenv("TORCHVISION_DATASETS_HOME")
+    if home is not None:
+        return pathlib.Path(home)
+
+    return HOME
diff --git a/torchvision/prototype/datasets/utils/__init__.py b/torchvision/prototype/datasets/utils/__init__.py
index ee45837a255..114e6507fdb 100644
--- a/torchvision/prototype/datasets/utils/__init__.py
+++ b/torchvision/prototype/datasets/utils/__init__.py
@@ -1,5 +1,3 @@
 from ._dataset import *
-from ._helpers import *
 from ._resource import *
-
 from . import _internal
diff --git a/torchvision/prototype/datasets/utils/_dataset.py b/torchvision/prototype/datasets/utils/_dataset.py
index 2b0b51623ea..2226fc7956f 100644
--- a/torchvision/prototype/datasets/utils/_dataset.py
+++ b/torchvision/prototype/datasets/utils/_dataset.py
@@ -100,12 +100,21 @@ def __init__(
         self,
         name: str,
         *,
+        categories: Union[int, Sequence[str], str, pathlib.Path],
         citation: Optional[str] = None,
         homepage: Optional[str] = None,
         license: Optional[str] = None,
         valid_options: Optional[Dict[str, Sequence[Any]]] = None,
     ) -> None:
         self.name = name.lower()
+
+        if isinstance(categories, int):
+            categories = [str(label) for label in range(categories)]
+        elif isinstance(categories, (str, pathlib.Path)):
+            with open(pathlib.Path(categories).expanduser().resolve(), "r") as fh:
+                categories = fh.readlines()
+        self.categories = categories
+
         self.citation = citation
         self.homepage = homepage
         self.license = license
diff --git a/torchvision/prototype/datasets/utils/_helpers.py b/torchvision/prototype/datasets/utils/_helpers.py
deleted file mode 100644
index 23df92b786e..00000000000
--- a/torchvision/prototype/datasets/utils/_helpers.py
+++ /dev/null
@@ -1,26 +0,0 @@
-import io
-from typing import Any, Tuple
-
-import numpy as np
-import PIL.Image
-
-__all__ = ["read_mat", "image_buffer_from_array"]
-
-
-def read_mat(file: io.BufferedIOBase, **kwargs: Any) -> Any:
-    try:
-        import scipy.io as sio
-    except ImportError as error:
-        raise ModuleNotFoundError(
-            "Package `scipy` is required to be installed to read .mat files."
-        ) from error
-
-    return sio.loadmat(file, **kwargs)
-
-
-def image_buffer_from_array(array: np.array, *, format: str) -> Tuple[str, io.BytesIO]:
-    image = PIL.Image.fromarray(array)
-    buffer = io.BytesIO()
-    image.save(buffer, format=format)
-    buffer.seek(0)
-    return f"tmp.{format}", buffer
diff --git a/torchvision/prototype/datasets/utils/_internal.py b/torchvision/prototype/datasets/utils/_internal.py
index af0b87e9be3..8cfbe40d7db 100644
--- a/torchvision/prototype/datasets/utils/_internal.py
+++ b/torchvision/prototype/datasets/utils/_internal.py
@@ -1,6 +1,40 @@
 import collections.abc
 import difflib
+import io
+import pathlib
+from typing import Any, Tuple
 from typing import Collection, Sequence, Callable
+from typing import Union
+
+import numpy as np
+import PIL.Image
+
+__all__ = [
+    "read_mat",
+    "image_buffer_from_array",
+    "sequence_to_str",
+    "add_suggestion",
+    "create_categories_file",
+]
+
+
+def read_mat(file: io.BufferedIOBase, **kwargs: Any) -> Any:
+    try:
+        import scipy.io as sio
+    except ImportError as error:
+        raise ModuleNotFoundError(
+            "Package `scipy` is required to be installed to read .mat files."
+        ) from error
+
+    return sio.loadmat(file, **kwargs)
+
+
+def image_buffer_from_array(array: np.array, *, format: str) -> Tuple[str, io.BytesIO]:
+    image = PIL.Image.fromarray(array)
+    buffer = io.BytesIO()
+    image.save(buffer, format=format)
+    buffer.seek(0)
+    return f"tmp.{format}", buffer
 
 
 def sequence_to_str(seq: Sequence, separate_last: str = "") -> str:
@@ -34,3 +68,10 @@ def add_suggestion(
         else alternative_hint(possibilities)
     )
     return f"{msg.strip()} {hint}"
+
+
+def create_categories_file(
+    root: Union[str, pathlib.Path], name: str, categories: Sequence[str]
+) -> None:
+    with open(root / f"{name}.categories", "w") as fh:
+        fh.write("\n".join(categories) + "\n")
diff --git a/torchvision/prototype/datasets/utils/_resource.py b/torchvision/prototype/datasets/utils/_resource.py
index 27d274c996a..3ea4a79ff0c 100644
--- a/torchvision/prototype/datasets/utils/_resource.py
+++ b/torchvision/prototype/datasets/utils/_resource.py
@@ -33,8 +33,9 @@ def __init__(self, url: str, *, sha256: str, file_name: str) -> None:
         self.file_name = file_name
 
     def to_datapipe(self, root: Union[str, pathlib.Path]) -> IterDataPipe:
+        path = (pathlib.Path(root) / self.file_name).expanduser().resolve()
         # FIXME
-        return FileLoader(IterableWrapper((str(pathlib.Path(root) / self.file_name),)))
+        return FileLoader(IterableWrapper((str(path),)))
 
 
 # TODO: add support for mirrors