Datasets documentation

Load image data

You are viewing v2.19.0 version. A newer version v2.20.0 is available.
Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

Load image data

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

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("beans", split="train")
>>> dataset[0]["image"]

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 general loading guide.

Local files

You can load a dataset from the image path. Use the 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("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'}


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:



Load your dataset by specifying imagefolder and the directory of your dataset in data_dir:

>>> from datasets import load_dataset

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

Load remote datasets from their URLs with the data_files parameter:

>>> dataset = load_dataset("imagefolder", data_files="", split="train")

Some datasets have a metadata file (metadata.csv/metadata.jsonl) associated with it, containing other information about the data like bounding boxes, text captions, and labels. The metadata is automatically loaded when you call load_dataset() and specify imagefolder.

To ignore the information in the metadata file, set drop_labels=False in load_dataset(), and allow ImageFolder to automatically infer the label name from the directory name:

>>> from datasets import load_dataset

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

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


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)