Deep Neural Networks

This repository contains Pytorch files that implement Basic Neural Networks for different datasets.
It includes various codes that implement variational autoencoder and generative adversial networks, along with implementation of some milestone papers in the field of Neural Networks, Riemannian Geometry and making use of this geometry along with Neural Networks to solve Computer Vision Problems.

Our work borrows motivation from the following papers :

1. Tutorial on VAE's
2. Riemannian Geometry of Deep Generative Model

Requirements

1. Python 2/3 (2 is preferred)
2. Numpy
3. Pytorch
4. Tensorboard
5. Matplotlib

Contents

1. Autoencoder: Contains files for implementing a basic Variational Autoencoder (VAE) of two types - generative and non-generative, for MNIST Dataset along with its results.
2. GANS: Contains files for implementing a basic Generative Adversarial Network along with its results.
3. NN_different_datasets: Contains files for implementing Neural Networks on different datasets (CIFAR10).
4. SavedModels: Contains saved models (.pt files) for different networks for various datasets.
5. plots: Contains files for various plots for interpreting purposes of different networks like loss plot, plotting class-wise mean and standard deviation etc.