Autoencoder with Fashion MNIST

This project implements an autoencoder to compress and reconstruct images from the Fashion MNIST dataset. The autoencoder is built using TensorFlow/Keras and visualizes the reconstruction quality. Additionally, the project calculates the Structural Similarity Index (SSIM) to evaluate the performance of the model.

Features

• Data preprocessing for the Fashion MNIST dataset.
• Autoencoder with dense layers for encoding and decoding.
• Visualization of original and reconstructed images.
• SSIM computation for evaluating reconstruction quality.

Prerequisites

Ensure you have the following packages installed:

• numpy
• matplotlib
• tensorflow
• scikit-image

You can install the necessary packages using pip:

pip install numpy matplotlib tensorflow scikit-image

Usage

1. Clone this repository:

   git clone https://github.com/KonyD/fashion-mnist-autoencoder.git
   cd fashion-mnist-autoencoder

2. Run the Python script:

   python autoencoder_fashion_mnist.py

3. The script will display:

   • Sample original and reconstructed images.
   • Average SSIM score printed in the console.

Code Overview

• Data Preprocessing:

  • Normalizes the Fashion MNIST images to a range of [0, 1].
  • Reshapes the dataset for compatibility with dense layers.

• Model Architecture:

  • Encoder: Compresses the input images into a lower-dimensional latent space.
  • Decoder: Reconstructs the images from the compressed latent space.

• Evaluation:

  • Visualizes original vs. reconstructed images.
  • Calculates SSIM scores for evaluating reconstruction quality.

Example Output

• Visualization of Original and Reconstructed Images:

  

• SSIM Score:

  SSIM:  0.7793266204555716  # Your results may vary

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgments

• Dataset: Fashion MNIST
• Libraries: TensorFlow, Keras, NumPy, Matplotlib, Scikit-Image

Feel free to contribute or raise issues if you encounter any problems!