# Conditional GANs For Painting Generation

## TL;DR

This paper explores the use of spectral normalization GANs for generating realistic oil painting images, introducing a novel conditional architecture for user-controlled face painting synthesis.

## Contribution

It compares different GAN variants for painting generation and proposes a new conditional GAN architecture for customizable face paintings.

## Key findings

- SN-GAN produced the most realistic paintings according to Sliced Wasserstein Distance.
- Spectral normalization improves the quality of generated paintings.
- The proposed conditional GAN allows user-specified face painting features.

## Abstract

We examined the use of modern Generative Adversarial Nets to generate novel images of oil paintings using the Painter By Numbers dataset. We implemented Spectral Normalization GAN (SN-GAN) and Spectral Normalization GAN with Gradient Penalty, and compared their outputs to a Deep Convolutional GAN. Visually, and quantitatively according to the Sliced Wasserstein Distance metric, we determined that the SN-GAN produced paintings that were most comparable to our training dataset. We then performed a series of experiments to add supervised conditioning to SN-GAN, the culmination of which is what we believe to be a novel architecture that can generate face paintings with user-specified characteristics.