ImageNette Dataset

The ImageNette dataset is a subset of ImageNet with 10 easily distinguishable classes, created by fast.ai to provide a quicker, easier-to-use version of ImageNet for software development and education. It contains 13,394 color images — 9,469 for training and 3,925 for validation — spanning classes such as tench, English springer, cassette player, and French horn, so image classification models train in minutes rather than the hours the full 1,000-class ImageNet requires.

Key Features

ImageNette contains 13,394 images across 10 classes: tench, English springer, cassette player, chain saw, church, French horn, garbage truck, gas pump, golf ball, and parachute.
It ships with a predefined split of 9,469 training and 3,925 validation color images of varying dimensions.
The 10 well-separated classes make ImageNette fast to train and easy to reason about, so it is widely used for prototyping and teaching image classification.

Dataset Structure

ImageNette ships with a predefined train/validation split, with each class stored in its own folder:

Split	Images	Classes
Train	9,469	10
Validation	3,925	10

The exact number of images varies per class, giving a realistic distribution for quick training and benchmarking without the scale of the full ImageNet dataset.

Applications

The ImageNette dataset is widely used for training and evaluating deep learning models in image classification tasks, such as Convolutional Neural Networks (CNNs) and other machine learning algorithms. Its straightforward format and well-chosen classes make it a handy resource for both beginner and experienced practitioners in machine learning and computer vision.

Usage

To train a model on the ImageNette dataset for 100 epochs with a standard image size of 224x224, you can use the following code snippets. For a comprehensive list of available arguments, refer to the model Training page.

!!! example "Train Example"

=== "Python"

    ```python
    from ultralytics import YOLO

    # Load a model
    model = YOLO("yolo26n-cls.pt")  # load a pretrained model (recommended for training)

    # Train the model
    results = model.train(data="imagenette", epochs=100, imgsz=224)
    ```

=== "CLI"

    ```bash
    # Start training from a pretrained *.pt model
    yolo classify train data=imagenette model=yolo26n-cls.pt epochs=100 imgsz=224
    ```

Sample Images and Annotations

The ImageNette dataset contains colored images of various objects and scenes, providing a diverse dataset for image classification tasks. Here are some examples of images from the dataset:

The example showcases the variety and complexity of the images in the ImageNette dataset, highlighting the importance of a diverse dataset for training robust image classification models.

ImageNette160 and ImageNette320

For faster prototyping and training, ImageNette is also available in two reduced sizes: ImageNette160 and ImageNette320. These datasets keep the same classes and structure as full ImageNette, but the images are resized to a smaller maximum edge length. They are particularly useful for preliminary model testing or when computational resources are limited.

To use these datasets, simply replace imagenette with imagenette160 or imagenette320 in the training command. The following code snippets illustrate this:

!!! example "Train Example with ImageNette160"

=== "Python"

    ```python
    from ultralytics import YOLO

    # Load a model
    model = YOLO("yolo26n-cls.pt")  # load a pretrained model (recommended for training)

    # Train the model with ImageNette160
    results = model.train(data="imagenette160", epochs=100, imgsz=160)
    ```

=== "CLI"

    ```bash
    # Start training from a pretrained *.pt model with ImageNette160
    yolo classify train data=imagenette160 model=yolo26n-cls.pt epochs=100 imgsz=160
    ```

!!! example "Train Example with ImageNette320"

=== "Python"

    ```python
    from ultralytics import YOLO

    # Load a model
    model = YOLO("yolo26n-cls.pt")  # load a pretrained model (recommended for training)

    # Train the model with ImageNette320
    results = model.train(data="imagenette320", epochs=100, imgsz=320)
    ```

=== "CLI"

    ```bash
    # Start training from a pretrained *.pt model with ImageNette320
    yolo classify train data=imagenette320 model=yolo26n-cls.pt epochs=100 imgsz=320
    ```

These smaller versions of the dataset allow for rapid iterations during development while still providing realistic image classification tasks. You can also manage classification datasets and run training in the cloud with Ultralytics Platform.

Citations and Acknowledgments

If you use the ImageNette dataset in your research or development work, please acknowledge it appropriately. For more information about ImageNette, visit the ImageNette dataset GitHub page.

We would like to acknowledge the fast.ai team for creating and maintaining ImageNette as a valuable resource for the machine learning and computer vision research community.

FAQ

What is the ImageNette dataset?

The ImageNette dataset is a simplified subset of ImageNet with 10 easily distinguishable classes such as tench, English springer, and French horn. It contains 13,394 images (9,469 training and 3,925 validation) and was created by fast.ai to offer a more manageable dataset for efficient training and evaluation of image classification models. This makes it especially useful for quick software development and educational purposes in machine learning and computer vision.

How many images and classes does ImageNette have?

ImageNette contains 13,394 images in total — 9,469 for training and 3,925 for validation — across 10 classes: tench, English springer, cassette player, chain saw, church, French horn, garbage truck, gas pump, golf ball, and parachute. Each class is stored in its own folder, following the standard classification layout Ultralytics expects.

How can I use the ImageNette dataset for training a YOLO model?

To train a YOLO model on ImageNette for 100 epochs, use the following commands. Make sure you have the Ultralytics YOLO environment set up.

!!! example "Train Example"

=== "Python"

    ```python
    from ultralytics import YOLO

    # Load a model
    model = YOLO("yolo26n-cls.pt")  # load a pretrained model (recommended for training)

    # Train the model
    results = model.train(data="imagenette", epochs=100, imgsz=224)
    ```

=== "CLI"

    ```bash
    # Start training from a pretrained *.pt model
    yolo classify train data=imagenette model=yolo26n-cls.pt epochs=100 imgsz=224
    ```

For more details, see the Training documentation page.

Why should I use ImageNette for image classification tasks?

The ImageNette dataset is advantageous for several reasons:

Quick and Simple: With only 10 classes and about 13,000 images, it is far less complex and time-consuming to train than the full ImageNet.
Educational Use: Ideal for learning and teaching the basics of image classification, since it requires less computational power and time.
Versatility: Widely used to train and benchmark image classification models, especially Convolutional Neural Networks (CNNs).

For more details on model training and dataset management, explore the Dataset Structure section.

Can the ImageNette dataset be used with different image sizes?

Yes, ImageNette is also available in two resized versions, ImageNette160 and ImageNette320, whose images are resized to smaller maximum edge lengths. These versions help with faster prototyping and are especially useful when computational resources are limited.