Image Denoising Using AutoEncoders in Keras and Python

Environment Setup:

To run the script, at least following required packages should be satisfied:

• Python
• Tensorflow
• NumPy
• MatplotLib

Dataset:

Dataset used here is standard MNIST Fashion Dataset, which comprises of 28 x 28 pixel images of 9 different fashion wears labelled from 0-9 as specified below:

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

Autoencoders:

Autoencoders are an unsupervised learning technique. It is an artificial neural network which performs task of data encoding. It typically comprises of 3 layers: Input,Hidden,Output. And has 4 components:

• Encoder
• Bottleneck Layer
• Decoder
• Reconstruction Loss

For Detailed Explaination, refer : Link

Output:

Denoising Images by adding 50% noise to training and testing data Input and Output Image comparision :

References:

• Rhyme - Project: Image denoising using Autoencoder
• Auto-Encoder: What Is It? And What Is It Used For? (Part 1)
• Variational Autoencoders are Beautiful