Words as Art Materials: Generating Paintings with Sequential GANs

TL;DR

This paper introduces a sequential GAN model that generates artistic images from keywords, utilizing advanced techniques to improve stability and producing diverse, high-resolution paintings evaluated through quantitative and user-based methods.

Contribution

It presents a novel sequential GAN architecture that processes keywords to generate artistic images with multiple styles, enhancing stability and diversity in outputs.

Findings

Generated diverse artistic images with various styles

Achieved competitive Fréchet Inception Distance scores

User study confirmed visual appeal and quality

Abstract

Converting text descriptions into images using Generative Adversarial Networks has become a popular research area. Visually appealing images have been generated successfully in recent years. Inspired by these studies, we investigated the generation of artistic images on a large variance dataset. This dataset includes images with variations, for example, in shape, color, and content. These variations in images provide originality which is an important factor for artistic essence. One major characteristic of our work is that we used keywords as image descriptions, instead of sentences. As the network architecture, we proposed a sequential Generative Adversarial Network model. The first stage of this sequential model processes the word vectors and creates a base image whereas the next stages focus on creating high-resolution artistic-style images without working on word vectors. To deal with the unstable nature of GANs, we proposed a mixture of techniques like Wasserstein loss, spectral normalization, and minibatch discrimination. Ultimately, we were able to generate painting images, which have a variety of styles. We evaluated our results by using the Fr\'echet Inception Distance score and conducted a user study with 186 participants.