Image Classification Using Convolutional Neural Networks (CNN) with Fashion MNIST Dataset

---

Problem Statement

The objective of this practical is to build a Convolutional Neural Network (CNN) to automatically classify grayscale images of clothing into one of ten fashion categories using the Fashion MNIST dataset.

---

Objective

• Understand the basic architecture of CNN.

• Load and preprocess the Fashion MNIST dataset.

• Build and train a CNN model.

• Evaluate the model's performance.

• Predict the class of unseen fashion images.

Dataset Description

Dataset Description

The Fashion MNIST dataset is a benchmark image classification dataset developed by Zalando Research. It contains 70,000 grayscale images of fashion products belonging to 10 different clothing categories. Each image has a resolution of 28 × 28 pixels and is represented as a grayscale image.

The original dataset consists of 70,000 images, which are pre-divided by TensorFlow into 60,000 training images and 10,000 testing images. This predefined split allows the CNN model to be trained on the training dataset and evaluated on the testing dataset without requiring manual data partitioning.  

Go through Jupyter Notebook

Dataset Summary

Attribute	Description
Dataset Name	Fashion MNIST
Total Images	70,000
Training Images	60,000
Testing Images	10,000
Image Size	28 × 28 pixels
Image Type	Grayscale
Number of Classes	10
Task	Multiclass Image Classification

---

Fashion MNIST Classes

Label	Category
0	T-shirt/Top
1	Trouser
2	Pullover
3	Dress
4	Coat
5	Sandal
6	Shirt
7	Sneaker
8	Bag
9	Ankle Boot