https://github.com/mehrab-kalantari/cnn-fashion-mnist

Building a Convolutional Neural Network on fashion MNIST dataset
https://github.com/mehrab-kalantari/cnn-fashion-mnist

cnn convolutional-neural-network convolutional-neural-networks deep-learning fashion-mnist

Last synced: 8 months ago
JSON representation

Building a Convolutional Neural Network on fashion MNIST dataset

• Host: GitHub
• URL: https://github.com/mehrab-kalantari/cnn-fashion-mnist
• Owner: Mehrab-Kalantari
• Created: 2023-09-07T15:17:41.000Z (about 2 years ago)
• Default Branch: main
• Last Pushed: 2023-09-07T15:31:03.000Z (about 2 years ago)
• Last Synced: 2025-01-16T09:42:31.085Z (10 months ago)
• Topics: cnn, convolutional-neural-network, convolutional-neural-networks, deep-learning, fashion-mnist
• Language: Jupyter Notebook
• Homepage:
• Size: 111 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# CNN on Fashion MNIST Dataset

[Dataset on kaggle](https://www.kaggle.com/datasets/zalando-research/fashionmnist)

## Contents

### Data Cleaning

In this part we just load and reshape the dataset

### Data Preprocessing

* Scaling

* Encoding

### Modeling

1. First block

    * Zero padding layer

    * Convolution layer with 32 kernels

    * Batch normalization layer

    * Dropout layer

    * Max pooling layer

2. Second block

   * Zero padding layer

   * Convolution layer with 64 kernels

   * Batch normalization layer

   * Dropout layer

   * Max pooling layer

3. Third block

   * Flatten layer

   * Dense layer with 128 units

   * Dropout layer

   * Dense layer with 10 units

We use adam optimizer and categorical cross entropy as loss

### Evaluation

* Accuracy plot

* ![p](sample/acc.png)

* Classification report (91% accuracy)

* Confusion matrix

* ![p2](sample/output.png)