DyArtbank: Diverse Artistic Style Transfer via Pre-trained Stable Diffusion and Dynamic Style Prompt Artbank

Zhanjie Zhang, Quanwei Zhang, Guangyuan Li, Junsheng Luan, Mengyuan Yang, Yun Wang, Lei Zhaocor1 · 2025

Read DyArtbank: Diverse Artistic Style Transfer via Pre-trained Stable Diffusion and Dynamic Style Prompt Artbank by Zhanjie Zhang, Quanwei Zhang, Guangyuan Li, Junsheng Luan, Mengyuan Yang, Yun Wang, Lei Zhaocor1 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-08392.txt.

Description

Zhanjie Zhang and six co-authors make style transfer produce variety instead of a single fixed look. Most tools restyle a photo after one artwork and return near-identical results each run, leaving users with little to enjoy. The team's framework, DyArtbank, leans on a pretrained image generator. A bank of learnable style prompts soaks up the look of a whole artwork collection and steers the generator on the fly. A second module feeds the generator strong content cues so the scene's fine structure survives the restyling.

Because the style bank can also sample fresh images from the learned look, it doubles as a source of extra training data. Tests on quality and on measured scores back the design. The authors post their code online. Note that the catalog garbles the last author's name; the paper's own byline reads Lei Zhao.

Catalog record

CREATORZhanjie Zhang, Quanwei Zhang, Guangyuan Li, Junsheng Luan, Mengyuan Yang, Yun Wang, Lei Zhaocor1
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE38 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-08392

Browse the full library · About the Tower of Babel