# Fashion-MNIST

## Facts

- Provider: Zalando Research (Han Xiao, Kashif Rasul, Roland Vollgraf)
- License: MIT (https://github.com/zalandoresearch/fashion-mnist/blob/master/LICENSE)
- Size: 4 files
- Format: Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/zalando-datasets/fashion_mnist
- Access: public, readable without an account
- Category: Machine Learning & AI
- Source: https://arxiv.org/abs/1708.07747
- Page: https://scigantic.com/archive/675b0a62-db5a-50fc-a0ba-d8e1a22e3df6
- Keywords: Machine Learning & AI, benchmark, Parquet, Hugging Face

## About

Fashion-MNIST is Zalando Research's benchmark of article images from the Zalando catalogue, released in 2017 as a direct drop-in replacement for the original MNIST handwritten digit dataset. It has a training set of 60,000 examples and a test set of 10,000 examples; each example is a 28x28 grayscale image associated with one of 10 clothing classes (T-shirt/top, Trouser, Pullover, Dress, Coat, Sandal, Shirt, Sneaker, Bag, Ankle boot). Because it shares MNIST's image size, split sizes and file structure, any pipeline written for MNIST runs on it unchanged, while being harder to solve, which is why it is a standard sanity-check and teaching set for image classification, convolutional networks, representation learning, active learning and dataset-shift experiments. Measured 2026-11-14: this Hugging Face copy holds the dataset as two Parquet files under fashion_mnist/ (train about 30.9 MB, test about 5.2 MB) with an image column and an integer label column; load with the datasets library (load_dataset('zalando-datasets/fashion_mnist')), pandas or pyarrow. The data and code are released by Zalando SE under the MIT License; cite Xiao, Rasul and Vollgraf, arXiv:1708.07747.

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("zalando-datasets/fashion_mnist")
print(ds)
```

## Data structure

File types in sampled files: Parquet

Example file: `fashion_mnist/test-00000-of-00001.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/675b0a62-db5a-50fc-a0ba-d8e1a22e3df6
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/675b0a62-db5a-50fc-a0ba-d8e1a22e3df6
- Over MCP: `get_schema_card` with id `675b0a62-db5a-50fc-a0ba-d8e1a22e3df6` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Benchmark datasets](https://scigantic.com/topics/benchmark)
- [Machine Learning & AI datasets](https://scigantic.com/topics/machine-learning-and-ai)
