This repository implements a basic autoencoder model using Keras to compress and reconstruct images from the Fashion MNIST dataset.
What is an Autoencoder?
An autoencoder is a type of unsupervised neural network that learns to represent input data in a compressed latent space. This compressed representation captures the essential features of the data while discarding noise or redundancy. By reconstructing the original data from the latent representation, the autoencoder can be used for dimensionality reduction, anomaly detection, and data visualization.
Project Structure:
encoder.py: The core Python script containing the code for loading data, building the autoencoder model, training, and visualizing results.
Running the Code:
Prerequisites: Install Python and required libraries,
Refer to requirements folder.
Download Data (if not using built-in Keras loading):
Download the Fashion MNIST dataset from http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/.
Place the downloaded files in the data directory (if created).
Run the script:
Execute python encoder.py in your terminal. Run :
Python3 encoder.py
Expected Output:
The script will train the autoencoder model and display a figure comparing 10 original Fashion MNIST images with their reconstructed versions.
Further Exploration:
Experiment with different network architectures (e.g., deeper layers, different activation functions). Visualize the latent space to explore how the model encodes different types of clothing. Use the autoencoder for dimensionality reduction on other image datasets. Feel free to contribute!
This is a basic implementation to get you started. We welcome contributions to improve the code, add visualizations, or explore other applications.