Metadata-Version: 2.4
Name: animal_mnist
Version: 0.5.5
Summary: Animal silhouette extension to MNIST - benchmarking dataset
Author-email: Sina Mohammadi <mohammadi.sina.q8@dc.tohoku.ac.jp>
License: MIT
Classifier: Programming Language :: Python :: 3
Classifier: Operating System :: OS Independent
Classifier: Intended Audience :: Science/Research
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: numpy
Requires-Dist: scikit-learn
Requires-Dist: tensorflow
Requires-Dist: pandas
Dynamic: license-file

# Animal-MNIST
Animal-MNIST is a dataset consisting of 10,000 28x28 grayscale images, mimicking the format of the MNIST dataset. There are 10 classes of animals (Namely Bear, Bird, Cat, Cow, Dog, Elephant, Giraffe, Horse, Sheep, Zebra) each consisting of 1000 pictures. This is intended to serve educational purposes and expand further on the MNIST - fashion MNIST benchmarking system for machine learning algorithms.

## Installation
```bash
pip install --upgrade animal-mnist
```

## Load the dataset

```python
from animal_mnist.dataset_Loader import load_animal_mnist_dataset

X, y = load_animal_mnist_dataset()
```


## Official train/validation/test split

```python
from animal_mnist.dataset_Loader import load_animal_mnist_splits

train_idx, val_idx, test_idx = load_animal_mnist_splits()

X_train, y_train = X[train_idx], y[train_idx]
X_val, y_val = X[val_idx], y[val_idx]
X_test, y_test = X[test_idx], y[test_idx]
```

The official class-stratified partition contains 7,200 training images, 800 validation images, and 2,000 test images. Split indices are fixed for reproducible comparisons.
