Communications of the ACM
Researchers Fine-Tune Control Over AI Image Generation
The new artificial intelligence method enables the system to create and retain a background image, while also creating figures that are consistent from picture to picture, but which show change or movement.
Credit: Tianfu Wu
Refined control over artificial intelligence (AI)-driven conditional image generation developed by North Carolina State University (NC State) researchers has potential for use in fields ranging from autonomous robotics to AI training.
NC State's Tianfu Wu said, "Like previous approaches, ours allows users to have the system generate an image based on a specific set of conditions. But ours also allows you to retain that image and add to it."
The approach also can rig specific components to be identifiably the same, but shifted position or somehow altered.
In testing the approach using the COCO-Stuff and the Visual Genome datasets, the technique bested previous state-of-the-art image generation methods.
Wu suggested applications for the technique like helping autonomous robots "imagine" the appearance of an end result before undertaking a given task, or producing images for AI training.
From NC State University News
View Full Article
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA
No entries found