leandrodevai/imagenette-320px-resplit

Imagenette 320px with Fixed Validation and Test Splits

Dataset Description

This dataset is a reproducible, Parquet-based version of the 320px configuration of frgfm/imagenette. Imagenette is a subset of ten readily classified ImageNet classes created for fast experimentation with image-classification methods.

This version preserves the source images, numeric labels, and label metadata. Its only data change is a fixed, stratified division of the original validation split into validation and test splits for image-classification and confidence-calibration experiments.

Dataset Sources

• Source dataset: frgfm/imagenette
• Original project: fastai/imagenette
• Image source: ImageNet
• Configuration: 320px
• License declared by the source dataset: Apache License 2.0

Dataset Structure

Each example contains:

{
    "image": PIL.Image.Image,
    "label": int,
}

The label feature is a ClassLabel with ten names.

Splits

Split	Examples	Intended use
train	9,469	Model training and training-only preprocessing statistics
validation	1,962	Model selection, early stopping, and calibration fitting
test	1,963	Final evaluation only
Total	13,394

The source training split remains unchanged. The source validation split contained 3,925 examples and was divided using a stratified 50/50 split with seed 42.

Label Metadata

The source metadata already matches Imagenette's ImageNet synset order. This derivative verifies and preserves the following mapping:

Label ID	ImageNet synset	Class name
0	n01440764	tench
1	n02102040	English springer
2	n02979186	cassette player
3	n03000684	chain saw
4	n03028079	church
5	n03394916	French horn
6	n03417042	garbage truck
7	n03425413	gas pump
8	n03445777	golf ball
9	n03888257	parachute

Dataset Creation

Split Procedure

The fixed held-out splits were created from the source validation split with:

validation_test = source["validation"].train_test_split(
    test_size=0.5,
    seed=42,
    stratify_by_column="label",
)

The resulting validation_test["train"] split became validation, and validation_test["test"] became test. Publishing these assignments as Parquet ensures that experiments use the same examples.

Preprocessing

The source 320px configuration resizes the shorter side of each image to 320 pixels while preserving its aspect ratio. Images are not guaranteed to be square.

No additional resizing, cropping, augmentation, or pixel normalization was applied while creating this dataset. The companion exploration notebook computes pixel-weighted RGB statistics from the complete training split only.

Usage

from datasets import load_dataset

dataset = load_dataset(
    "leandrodevai/imagenette-320px-resplit",
    "320px",
)

train_dataset = dataset["train"]
validation_dataset = dataset["validation"]
test_dataset = dataset["test"]

For calibration experiments:

1. Train the classifier on train.
2. Fit temperature scaling or another post-hoc calibrator on validation.
3. Report classification and calibration metrics once on test.

The test split should not be used for architecture selection, hyperparameter tuning, early stopping, or calibrator fitting.

Intended Uses

This dataset is suitable for:

• supervised image classification;
• confidence-calibration and reliability-diagram experiments;
• evaluation of expected calibration error, negative log-likelihood, and Brier score;
• post-hoc calibration methods such as temperature scaling;
• controlled comparisons between pretrained and from-scratch training strategies.

Limitations

• The dataset contains only ten classes and does not represent general image classification.
• Images inherit ImageNet collection biases, including correlations between subjects, backgrounds, framing, and photographic style.
• The validation and test sets originate from the same source split and should not be interpreted as independent data-collection domains.
• Calibration results on this dataset may not transfer to other datasets, domain shifts, corruptions, or deployment settings.

Personal and Sensitive Information

As with the source ImageNet data, some images may contain people, locations, text, or other contextual information. No additional personal information was added during preparation.

Licensing

The source Hugging Face dataset declares the Apache License 2.0. This derivative preserves that declaration. Users are responsible for verifying that their intended use complies with the source dataset, ImageNet terms, and any rights associated with individual images.

Citation

If this dataset is useful, cite the original Imagenette project:

@software{Howard_Imagenette_2019,
  title = {Imagenette: A smaller subset of 10 easily classified classes from Imagenet},
  author = {Jeremy Howard},
  year = {2019},
  month = {March},
  publisher = {GitHub},
  url = {https://github.com/fastai/imagenette}
}

When reporting experiments, also document that this fixed-split derivative uses a stratified 50/50 division of the original validation split with seed 42.