Autoencoders are really emerging as solutions to many problems like image reconstruction, image denoising, dimensionality reduction, etc. One of the recent break through in autoencoder society is the invention of GANs, generative adversal networks can be used to produce novel data, hence from this all we can say autoencoders are the solution for many huge problems and issues mankind is having. WE used a 4 layered autoencoder to remove the noise in an image, the neurons used are 784-300-150-300-784.
Dataset we used is simple old classic MNIST, the reason was ease of computation and fast training of the model.