Autoencoders is a system of Neural Networks (NN) that have the same input an output. Autoencoders cosists of 2 NN, the encoder and decoder. This project is to demonstrate the implementation of autoencoders when generating image and audio.
For image generation, I used the MNIST dataset. For audio, I used FSDD.
Original Input written numbers (1st row) vs Generated Output written numbers (2nd row)
- var_autoencoder.py - Variational autoencoder module, Made out of 2 mirrored convolutional neural networks
- preprocessing_pipeline.py - Module for preprocessing data
- generator.py - Module for generating data