Convolutional Variational Autoencoder (VAE)

This repository contains an implementation of a Convolutional Variational Autoencoder (VAE) using PyTorch. The project trains the VAE on the CelebA dataset and generates face images.

Features

• Convolution Variational Autoencoder implementation in PyTorch
• Trained on CelebA dataset
• Customizable hyperparameters for flexibility
• Progress tracking with tqdm
• Possible to specify parameter "attribute" in order to generate images with one of the 40 specified attributes form the dataset.

Installation

0. This project is coded in Python. For installation, please consult: https://www.python.org/. If Python version is lower than 3.14, pip might need to be installed separately. Try the following command, or visit https://pypi.org/project/pip/.

   python -m ensurepip --upgrade
   

1. Clone the repository:

   git clone https://github.com/santysangro/Convolutional-Variational-Autoencoder.git
   cd Convolutional-Variational-Autoencoder  
   pip install -r requirements.txt

Running

To train the model:

   python scripts/train.py 

To generate images with specified attributes:

   python scripts/inference.py --num-images 5 --attribute Similing

Dataset

Download the CelebA dataset from this link and place it in the dataset/ directory.