Answers are grounded in this paper and cite their sources.
Computational framework demonstrates enhanced aesthetic quality and semantic consistency in AI-generated Chinese paintings, highlighting improved capture of traditional brushwork and artistic spirit.
While the current AIGC (Artificial Intelligence Generated Content) generation of Chinese paintings can reproduce the forms of ink and wash, it generally lacks the spirit of brush and ink and aesthetic conception, resulting in results that remain at the level of visual imitation. To address the lack of artistic spirit in AIGC, this paper constructs a generation system integrating brush and ink essence. It begins by collecting and semantically annotating a dataset with brushstrokes, white space, and artistic elements. CLIP multimodal encoding establishes image-to-semantic mapping. An a priori knowledge-based brush and ink control module is embedded within the Stable Diffusion framework to constrain generation, enhancing control over brushstroke flow and ink gradation. Philosophical concepts like virtuality/reality and white space are embedded using knowledge graphs. Interactive artist feedback is incorporated for collaborative creation. Experiments show significant improvements over the baseline: SSIM of 0.87 and FID of 18.5, a 40.3% improvement in overall aesthetic score, and a 26.2% improvement in semantic consistency CLIPScore. This confirms that the method effectively enhances the clarity, style, authenticity, and traditional artistic expression of generated Chinese paintings, while also supporting controllable visual pattern generation.
No takes yet. Share an insight, caveat, or question.
E. B. Fan (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: