Fashion-MNIST Dataset

Fashion-MNIST is a dataset of Zalando's article images designed as a drop-in replacement for the original MNIST dataset. It's ideal for quick prototyping, learning, and benchmarking classification algorithms.

Overview

Dataset Statistics:

• 60,000 training examples
• 10,000 test examples
• 10 classes: T-shirt/top, Trouser, Pullover, Dress, Coat, Sandal, Shirt, Sneaker, Bag, Ankle boot
• 28×28 grayscale images
• ~60MB total size (downloads quickly)

Class Labels:

• 0: T-shirt/top
• 1: Trouser
• 2: Pullover
• 3: Dress
• 4: Coat
• 5: Sandal
• 6: Shirt
• 7: Sneaker
• 8: Bag
• 9: Ankle boot

Reference: Fashion-MNIST GitHub

Installation

Fashion-MNIST requires no additional dependencies beyond the core OpenTryOn package. The dataset will be automatically downloaded on first use.

Usage

Class-Based Approach (Recommended)

from tryon.datasets import FashionMNIST

# Create dataset instance (downloads automatically)
dataset = FashionMNIST(download=True)

# Get dataset info
info = dataset.get_info()
print(f"Train size: {info['train_size']}")  # 60000
print(f"Test size: {info['test_size']}")    # 10000
print(f"Classes: {info['num_classes']}")    # 10

# Load the dataset
(train_images, train_labels), (test_images, test_labels) = dataset.load(
    normalize=True,
    flatten=False
)

print(f"Training set: {train_images.shape}")  # (60000, 28, 28)
print(f"Test set: {test_images.shape}")        # (10000, 28, 28)

# Get class names
print(dataset.get_class_name(0))  # 'T-shirt/top'
print(dataset.get_class_names())  # ['T-shirt/top', 'Trouser', ...]

Function-Based Approach

from tryon.datasets import load_fashion_mnist, get_fashion_mnist_class_name

# Load dataset
(train_images, train_labels), (test_images, test_labels) = load_fashion_mnist(
    normalize=True,
    flatten=False
)

# Get class name
class_name = get_fashion_mnist_class_name(0)  # 'T-shirt/top'

API Reference

Class: FashionMNIST

Fashion-MNIST dataset adapter class.

Constructor

FashionMNIST(data_dir: Optional[str] = None, download: bool = True)

Parameters:

• data_dir (str, optional): Directory to store the dataset. Defaults to ~/.opentryon/datasets/fashion_mnist
• download (bool): If True, download the dataset if it doesn't exist. Default: True

Example:

# Use default directory
dataset = FashionMNIST(download=True)

# Use custom directory
dataset = FashionMNIST(
    data_dir="./my_datasets/fashion_mnist",
    download=True
)

Methods

load(normalize=False, flatten=False)

Load Fashion-MNIST dataset into memory.

Parameters:

• normalize (bool): If True, normalize pixel values to [0, 1] range. Default: False
  • When False: pixel values are integers 0-255
  • When True: pixel values are floats 0.0-1.0
• flatten (bool): If True, flatten images to 1D arrays. Default: False
  • When False: images are shape (28, 28)
  • When True: images are shape (784,)

Returns:

• (train_data, test_data) tuple where each is (images, labels)
  • train_images: numpy array of shape (60000, 28, 28) or (60000, 784) if flatten=True
  • train_labels: numpy array of shape (60000,) with integer values 0-9
  • test_images: numpy array of shape (10000, 28, 28) or (10000, 784) if flatten=True
  • test_labels: numpy array of shape (10000,) with integer values 0-9

Data Types:

• Images: uint8 if normalize=False, float32 if normalize=True
• Labels: uint8

Example:

dataset = FashionMNIST()

# Load without normalization
(train_images, train_labels), (test_images, test_labels) = dataset.load(
    normalize=False,
    flatten=False
)
print(train_images.dtype)  # uint8
print(train_images.shape)  # (60000, 28, 28)

# Load with normalization
(train_images, train_labels), (test_images, test_labels) = dataset.load(
    normalize=True,
    flatten=False
)
print(train_images.dtype)  # float32
print(train_images.min(), train_images.max())  # 0.0 1.0

# Load flattened
(train_images, train_labels), (test_images, test_labels) = dataset.load(
    normalize=True,
    flatten=True
)
print(train_images.shape)  # (60000, 784)

get_info()

Get dataset information and metadata.

Returns:

• dict: Dictionary containing:
  • name: Dataset name ("Fashion-MNIST")
  • train_size: Number of training examples (60000)
  • test_size: Number of test examples (10000)
  • num_classes: Number of classes (10)
  • image_shape: Image shape tuple (28, 28)
  • normalized: Whether data is normalized (updated after load())
  • flattened: Whether data is flattened (updated after load())

Example:

dataset = FashionMNIST()
info = dataset.get_info()
print(info)
# {
#     'name': 'Fashion-MNIST',
#     'train_size': 60000,
#     'test_size': 10000,
#     'num_classes': 10,
#     'image_shape': (28, 28),
#     'normalized': False,
#     'flattened': False
# }

get_class_name(class_id: int)

Get the name of a class by its ID.

Parameters:

• class_id (int): Class ID (0-9)

Returns:

• str: Class name

Example:

dataset = FashionMNIST()
print(dataset.get_class_name(0))  # 'T-shirt/top'
print(dataset.get_class_name(9))  # 'Ankle boot'

get_class_names()

Get all class names.

Returns:

• list[str]: List of all class names

Example:

dataset = FashionMNIST()
class_names = dataset.get_class_names()
print(class_names)
# ['T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat',
#  'Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle boot']

Functions

load_fashion_mnist(normalize=False, flatten=False)

Convenience function to load Fashion-MNIST dataset.

Parameters:

• normalize (bool): Normalize pixel values to [0, 1]. Default: False
• flatten (bool): Flatten images to 1D arrays. Default: False

Returns:

• (train_data, test_data) tuple where each is (images, labels)

Example:

from tryon.datasets import load_fashion_mnist

(train_images, train_labels), (test_images, test_labels) = load_fashion_mnist(
    normalize=True,
    flatten=False
)

get_fashion_mnist_class_name(class_id: int)

Get class name by ID.

Parameters:

• class_id (int): Class ID (0-9)

Returns:

• str: Class name

get_fashion_mnist_class_names()

Get all class names.

Returns:

• list[str]: List of all class names

Best Practices

Normalization

For machine learning models, it's recommended to normalize pixel values:

dataset = FashionMNIST()
(train_images, train_labels), (test_images, test_labels) = dataset.load(
    normalize=True,  # Normalize to [0, 1]
    flatten=False
)

Memory Management

Fashion-MNIST is small enough (~60MB) to load entirely into memory. However, if you're working with limited memory:

# Load only what you need
dataset = FashionMNIST()
(train_images, train_labels), _ = dataset.load(normalize=True)
# Use only training set, ignore test set

Visualization

import matplotlib.pyplot as plt
from tryon.datasets import FashionMNIST

dataset = FashionMNIST()
(train_images, train_labels), (test_images, test_labels) = dataset.load(
    normalize=False,
    flatten=False
)

# Visualize a sample
plt.figure(figsize=(10, 10))
for i in range(25):
    plt.subplot(5, 5, i + 1)
    plt.imshow(train_images[i], cmap='gray')
    plt.title(dataset.get_class_name(train_labels[i]))
    plt.axis('off')
plt.tight_layout()
plt.show()

Examples

See Dataset Examples for complete usage examples.

Troubleshooting

Download Issues

If the dataset fails to download:

1. Check your internet connection
2. Verify write permissions to the data directory
3. Try specifying a custom data_dir:

   dataset = FashionMNIST(data_dir="./my_datasets/fashion_mnist", download=True)

Memory Issues

Fashion-MNIST is small, but if you encounter memory issues:

• Use normalize=False to keep data as uint8 (smaller memory footprint)
• Load only the split you need (train or test)