A new AI framework called Text2Poster automatically generates visually effective posters from textual information.

Posters have been widely used in numerous commercial and non-profit contexts to promote and disseminate information as a type of media with both artistic and practical elements. For example, e-commerce companies use flashy banners to advertise their products. Social event websites, such as conference ones, are often adorned with opulent and educational banners. These high-quality banners are created by integrating stylized lettering into appropriate background images, which requires a lot of manual editing and non-quantitative aesthetic intuition. However, such a slow and subjective approach cannot meet the enormous and rapidly growing demand for well-designed signs in real-world applications, diminishing the effectiveness of information dissemination and resulting in less-than-ideal marketing effects.

In this work, they offer Text2Poster, a unique data-driven framework that produces an effective automatic poster generator. Text2Poster initially uses a considerable pretrained visual textual model to retrieve appropriate background images from the input texts, as seen in the figure below. The framework then samples the intended layout layout to set the layout of the texts, then repeatedly refines the layout using cascading autoencoders. Finally, it gets the text color and font from a collection of colors and fonts that include semantic tags. They acquire the modules of the framework by using self-monitored and weak learning techniques. Experiments show that their Text2Poster system can automatically produce high-quality posters, outperforming academic and commercial rivals on both objective and subjective metrics.

The stages that the backend takes are the following:

1. Using a trained visual-textual model to retrieve images: They are interested in investigating photos that are “weakly associated” with sentences while collecting background images for poster development. For example, they like to discover images with love metaphors when they collect photos for the term “Bob and Alice’s Wedding,” such as an image of a white church against a blue sky. They use the BriVL, one of SOTA’s pretrained visual and textual models, to achieve this goal by retrieving background images from the texts.
2. Using cascading autoencoders for pattern prediction, smooth sections of the image are first found. Once the smooth areas are found, the smooth region is colored on the salience map. An estimated amplifier design distribution is now presented.
3. Text styling: The text is combined with the original image based on the intended arrangement.

They have a GitHub page where you can access the inference code to use Text2Poster. Download the source code files to run the program. Another way to use the program is to use its quickstart APIs. All usage details are written on their GitHub page.

review the Paper Y Github. All credit for this research goes to the researchers of this project. Also, don’t forget to join our reddit page, discord channel, Y electronic newsletterwhere we share the latest AI research news, exciting AI projects, and more.