Text2Poster: Laying out Stylized Texts on Retrieved Images

TL;DR

Text2Poster is a data-driven framework that automatically generates stylized posters from text by retrieving images, laying out texts iteratively, and stylizing them, reducing manual effort and artistic skill requirements.

Contribution

It introduces a novel, weakly- and self-supervised framework for automatic poster generation that combines image retrieval, iterative text layout, and stylization.

Findings

Outperforms state-of-the-art methods in poster quality

Effective use of weakly- and self-supervised learning strategies

Demonstrates superiority over existing academic and commercial tools

Abstract

Poster generation is a significant task for a wide range of applications, which is often time-consuming and requires lots of manual editing and artistic experience. In this paper, we propose a novel data-driven framework, called \textit{Text2Poster}, to automatically generate visually-effective posters from textual information. Imitating the process of manual poster editing, our framework leverages a large-scale pretrained visual-textual model to retrieve background images from given texts, lays out the texts on the images iteratively by cascaded auto-encoders, and finally, stylizes the texts by a matching-based method. We learn the modules of the framework by weakly- and self-supervised learning strategies, mitigating the demand for labeled data. Both objective and subjective experiments demonstrate that our Text2Poster outperforms state-of-the-art methods, including academic research and commercial software, on the quality of generated posters.