Research of Image Generation Based on Knowledge Graph

In recent years, text-to-image generation technology has achieved significant breakthroughs, especially with the advancement of Diffusion models. Sometimes the model's lack of knowledge or insufficient understanding of the text prompts to generate some unreasonable images. This paper aims to explore how combining knowledge graphs with text prompts can enhance the image generation process by providing additional relevant knowledge about the text prompts, thereby improving the capability of image generation without retraining the diffusion model.