Load image data

Image datasets have [Image] type columns, which contain PIL objects.

[!TIP] To work with image datasets, you need to have the vision dependency installed. Check out the installation guide to learn how to install it.

When you load an image dataset and call the image column, the images are decoded as PIL Images:

>>> from datasets import load_dataset, Image

>>> dataset = load_dataset("AI-Lab-Makerere/beans", split="train")
>>> dataset[0]["image"]

[!WARNING] Index into an image dataset using the row index first and then the image column - dataset[0]["image"] - to avoid decoding and resampling all the image objects in the dataset. Otherwise, this can be a slow and time-consuming process if you have a large dataset.

For a guide on how to load any type of dataset, take a look at the <a class="underline decoration-sky-400 decoration-2 font-semibold" href="./loading">general loading guide</a>.

Local files

You can load a dataset from the image path. Use the [~Dataset.cast_column] function to accept a column of image file paths, and decode it into a PIL image with the [Image] feature:

>>> from datasets import Dataset, Image

>>> dataset = Dataset.from_dict({"image": ["path/to/image_1", "path/to/image_2", ..., "path/to/image_n"]}).cast_column("image", Image())
>>> dataset[0]["image"]
<PIL.PngImagePlugin.PngImageFile image mode=RGBA size=1200x215 at 0x15E6D7160>]

If you only want to load the underlying path to the image dataset without decoding the image object, set decode=False in the [Image] feature:

>>> dataset = load_dataset("AI-Lab-Makerere/beans", split="train").cast_column("image", Image(decode=False))
>>> dataset[0]["image"]
{'bytes': None,
 'path': '/root/.cache/huggingface/datasets/downloads/extracted/b0a21163f78769a2cf11f58dfc767fb458fc7cea5c05dccc0144a2c0f0bc1292/train/bean_rust/bean_rust_train.29.jpg'}

ImageFolder

You can also load a dataset with an ImageFolder dataset builder which does not require writing a custom dataloader. This makes ImageFolder ideal for quickly creating and loading image datasets with several thousand images for different vision tasks. Your image dataset structure should look like this:

folder/train/dog/golden_retriever.png
folder/train/dog/german_shepherd.png
folder/train/dog/chihuahua.png

folder/train/cat/maine_coon.png
folder/train/cat/bengal.png
folder/train/cat/birman.png

Alternatively it should have metadata, for example:

folder/train/metadata.csv
folder/train/0001.png
folder/train/0002.png
folder/train/0003.png

If the dataset follows the ImageFolder structure, then you can load it directly with [load_dataset]:

>>> from datasets import load_dataset

>>> dataset = load_dataset("username/dataset_name")
>>> # OR locally:
>>> dataset = load_dataset("/path/to/folder")

For local datasets, this is equivalent to passing imagefolder manually in [load_dataset] and the directory in data_dir:

>>> dataset = load_dataset("imagefolder", data_dir="/path/to/folder")

Then you can access the videos as PIL.Image objects:

>>> dataset["train"][0]
{"image": <PIL.PngImagePlugin.PngImageFile image mode=RGBA size=1200x215 at 0x15E6D7160>, "label": 0}

>>> dataset["train"][-1]
{"image": <PIL.PngImagePlugin.PngImageFile image mode=RGBA size=1200x215 at 0x15E8DAD30>, "label": 1}

To ignore the information in the metadata file, set drop_metadata=True in [load_dataset]:

>>> from datasets import load_dataset

>>> dataset = load_dataset("username/dataset_with_metadata", drop_metadata=True)

If you don't have a metadata file, ImageFolder automatically infers the label name from the directory name. If you want to drop automatically created labels, set drop_labels=True. In this case, your dataset will only contain an image column:

>>> from datasets import load_dataset

>>> dataset = load_dataset("username/dataset_without_metadata", drop_labels=True)

Finally the filters argument lets you load only a subset of the dataset, based on a condition on the label or the metadata. This is especially useful if the metadata is in Parquet format, since this format enables fast filtering. It is also recommended to use this argument with streaming=True, because by default the dataset is fully downloaded before filtering.

>>> filters = [("label", "=", 0)]
>>> dataset = load_dataset("username/dataset_name", streaming=True, filters=filters)

[!TIP] For more information about creating your own ImageFolder dataset, take a look at the Create an image dataset guide.

WebDataset

The WebDataset format is based on a folder of TAR archives and is suitable for big image datasets. Because of their size, WebDatasets are generally loaded in streaming mode (using streaming=True).

You can load a WebDataset like this:

>>> from datasets import load_dataset

>>> dataset = load_dataset("webdataset", data_dir="/path/to/folder", streaming=True)

Lance

Lance is an open multimodal lakehouse table format. Lance tables can natively store not only text and scalar values, but also large binary objects (blobs) such as images, audio, and video alongside your tabular data.

Lance keeps your metadata and image blobs together in one place, while still letting you efficiently scan only the metadata columns you care about without loading image bytes. When you're ready, you can fetch a small subset of rows (including the image blobs) and write them directly to files on your local filesystem.

from datasets import load_dataset

# Return as a Hugging Face dataset
ds = load_dataset(
    "lance-format/laion-1m",
    split="train",
    streaming=True
)

dir_name = "laion_samples"
Path(dir_name).mkdir(exist_ok=True)

for idx, row in enumerate(ds.take(3)):
    with open(f"{dir_name}/{idx}.jpg", "wb") as f:
        f.write(row["image"])

In this example, the image column contains the encoded image bytes, so you can write them directly to .jpg files.

[!NOTE] The datasets API doesn't currently push down operations to the Lance table, so for larger datasets it may be slow. For now, you'll get much better performance using the lance Python package directly. See the documentation on the Hub for examples on usage.

Image decoding

By default, images are decoded sequentially as PIL.Images when you iterate on a dataset. However it is possible to speed up the dataset significantly using multithreaded decoding:

>>> import os
>>> num_threads = num_threads = min(32, (os.cpu_count() or 1) + 4)
>>> dataset = dataset.decode(num_threads=num_threads)
>>> for example in dataset:  # up to 20 times faster !
...     ...

You can enable multithreading using num_threads. This is especially useful to speed up remote data streaming. However it can be slower than num_threads=0 for local data on fast disks.

If you are not interested in the images decoded as PIL.Images and would like to access the path/bytes instead, you can disable decoding:

>>> dataset = dataset.decode(False)

Note: [IterableDataset.decode] is only available for streaming datasets at the moment.