Implementation of Autoencoder Denoising

The website: 

 Building Autoencoders in Keras

 under section “Application to image denoising” describes a denoising autoencoder. 

The  program described receives as input and denoises images of the mnist digits database.  

a. The first task is to put together the program of this autoencoder that receives and  plots the same input and produce and plot the same output images as those  described in the website. Then, extend this program in the ways described below: 

b. Extend this denoising autoencoder to receive and denoise the color images of the  CIFAR100 dataset available from Keras. 

c. Make this autoencoder deeper. Does this improve the performance? 

d. (Extra credit) As described at the end of the section, you can scale this process to a  bigger convnet. 

For example, you can build a program for document denoising.  Kaggle has an interesting dataset  Denoising Dirty Documents | Kaggle 

Hand in a different Python program for each of a, b, c, and if you choose d as well

Answer rating: 100% (QA)

Here are the Python programs for each of a b c and d a Denoising MNIST digits Python import kerasfrom kerasdatasets import mnistfrom keraslayers import Input Dense Conv2D MaxPooling2D Flatten Reshape ... View the full answer