Article: Application of knowledge graph-enhanced generative diffusion model for brand visual generation Journal: International Journal of Information and Communication Technology (IJICT) 2025 Vol.26 No.51 pp.50 - 69 Abstract: As brand visual content becomes more important in marketing, it has gotten harder to make images that fit the brand's identity. This research presents a generative diffusion approach utilising knowledge graph improvement for brand visual generation to tackle this issue. By putting the brand knowledge graph into the generation process, the system gives semantic direction for making images. The first step in the plan is to construct a knowledge graph that shows the brand's main qualities. Then, a generative diffusion model based on the knowledge graph is made and tested to see how well it works for brand visual production. The model enhances the inception score (IS) by 21.3% and diminishes the Fréchet inception distance (FID) by 19.5% in comparison to the conventional generative model. The model makes pictures that are consistent with the brand and seem beautiful, with good brand customisation and innovation. Inderscience Publishers - linking academia, business and industry through research

Title: Application of knowledge graph-enhanced generative diffusion model for brand visual generation

Authors: Fengfeng Shi

Addresses: College of Cultural Creativity and Tourism, Yuncheng Vocational and Technical University, Yuncheng 044000, China

Abstract: As brand visual content becomes more important in marketing, it has gotten harder to make images that fit the brand's identity. This research presents a generative diffusion approach utilising knowledge graph improvement for brand visual generation to tackle this issue. By putting the brand knowledge graph into the generation process, the system gives semantic direction for making images. The first step in the plan is to construct a knowledge graph that shows the brand's main qualities. Then, a generative diffusion model based on the knowledge graph is made and tested to see how well it works for brand visual production. The model enhances the inception score (IS) by 21.3% and diminishes the Fréchet inception distance (FID) by 19.5% in comparison to the conventional generative model. The model makes pictures that are consistent with the brand and seem beautiful, with good brand customisation and innovation.

Keywords: knowledge graph; generative diffusion model; brand visual generation; brand consistency; visual quality.

DOI: 10.1504/IJICT.2025.151077

International Journal of Information and Communication Technology, 2025 Vol.26 No.51, pp.50 - 69

Received: 26 Aug 2025
Accepted: 13 Nov 2025
Published online: 12 Jan 2026 *

Title: Application of knowledge graph-enhanced generative diffusion model for brand visual generation

Keep up-to-date