Generative Models

This repository explores various data augmentation techniques and image generation models, such as GAN, WGAN, CGAN, VAE, and CVAE using datasets like MNIST and FashionMNIST. It includes Jupyter notebooks with step-by-step implementations and reusable source code to facilitate exploration and understanding of augmentation techniques. In addition, in the folder "tabular_data_aug" we explore several augmentation tecniques for tabular data, such as SMOTE, CT-GAN, T-VAE and GReaT.

Repository Structure

  data_augmentation/ 
  │
  ├── notebooks/                
  │   ├── wgan_fashion_mnist.ipynb 
  │   ├── gan_mnist.ipynb 
  │   └── logo_sapienza_augmentation.ipynb  
  │
  ├── src/                      
  │   ├── data_loader.py        
  │   ├── models/                
  │   │   ├── gan_mnist.py  
  │   │   ├── gan_fashion_mnist.py   
  │   │   ├── wgan_fashion_mnist.py           
  │   │   ├── cgan_mnist.py   
  │   │   ├── cgan_fashion_mnist.py  
  │   │   ├── vae_mnist.py 
  │   │   ├── vae_fashion_mnist.py 
  │   │   ├── cvae_mnist.py 
  │   │   └── cvae_fashion_mnist.py  
  │   ├── main_gan.py   
  │   ├── main_vae.py  
  │   └── utils/                  
  │       └── visualization.py  
  │
  ├── photos/
  │   ├── Sapienza_logos/    
  │   ├── gan/            
  │   ├── wgan/ 
  │   ├── cgan/ 
  │   └── vae/  
  │
  │
  ├── tabular_data_aug/                       
  │   │── src/ 
  │   │   ├── data_loader.py        
  │   │   ├── models/
  │   │   │   └── mlp_classifier.py/  
  │   │   ├── main_SMOTE.py 
  │   │   ├── main_CTGAN.py 
  │   │   ├── main_TVAE.py 
  │   │   └── main_GREAT.py  
  │   │── images/ 
  │   └── README.md  
  │              
  ├── README.md                 
  └── requirements.txt                 

MNIST experiment

GAN

CGAN

VAE

CVAE

FashionMNIST experiment

GAN

WGAN

CGAN

VAE

CVAE