Variational Autoencoder (VAE) Project

This repository contains the implementation of a Variational Autoencoder (VAE) for learning latent representations of data. The project includes data preprocessing, model training, and visualization of the latent space.

Table of Contents

• Introduction
• Prerequisites
• Installation
• Usage
• Project Structure
• Results

Introduction

Variational Autoencoders (VAEs) are generative models that learn the underlying distribution of data in a latent space. This project demonstrates the use of VAEs to encode data into a lower-dimensional latent space and visualize the learned representations.

Prerequisites

• Python 3.7 or higher
• PyTorch
• torchvision
• matplotlib
• scikit-learn
• numpy

Installation

1. Clone the repository:

   git clone https://github.com/yourusername/vae-project.git
   cd vae-project
   

2. Install dependencies:

   pip install -r requirements.txt

Usage

Project Structure

vae-project/ │ ├── dataset.py # Defines SingleDataset class ├── vae.py # Defines VAE model ├── train.py # VAE training script ├── plotting_2D_latent_space.py # Visualizes 2D latent space ├── weights/ # Directory for model checkpoints │ ├── lat_1.ckpt # Checkpoint for 1D VAE │ ├── lat_2.ckpt # Checkpoint for 2D VAE │ ├── lat_5.ckpt # Checkpoint for 4D VAE │ ├── lat_10.ckpt # Checkpoint for 10D VAE │ ├── lat_25.ckpt # Checkpoint for 25D VAE │ └── lat_50.ckpt # Checkpoint for 50D VAE ├── requirements.txt # Python package dependencies ├── image-to-z-4bar.ipynb # Converts 4-bar CC images to latent representation ├── image-to-z-6bar.ipynb # Converts 6-bar CC images to latent representation ├── image-to-z-8bar.ipynb # Converts 8-bar CC images to latent representation ├── BSIdict_468.json # Kinematic structure information for mechs ├── KV_468.json # Mech information with enumeration ├── VK_468.json # Mech information with enumeration ├── metrics.py # Calculates Chamfer and Ordered distances ├── normalize.py # Normalization script ├── path_decomposition.py # Simulation file for path decomposition ├── server.py # Headless server script ├── simulator/ # Headless server folder │ └── headless_main_accuracy.py # Calculates accuracy for test dataset │ └── headless_main_novelty.py # Calculates novelty for test dataset └── README.md # Project readme file