Generation of Dunhuang Murals Using Emotion Vectors and Style Matrix Control
International Journal of Computational Intelligence Systems · 25.04.2026 · 19 (1)
DOI: 10.1007/s44196-026-01346-4 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Computational Intelligence Systems, online erschienen am 25.04.2026. Verfasst von Yuchen Song. Die Arbeit nennt 33 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).
Abstract (Originalsprache)
Emotion-driven generation of Dunhuang murals remains challenging because affect, visual style, and iconography must remain coherent within a historically grounded aesthetic. We propose an emotion–style collaborative framework built on latent diffusion, where a multimodal encoder fuses text and speech into a 128-dimensional emotion vector and an LSTM-based pattern estimator captures motif periodicity. These signals are mapped to four interpretable style parameters (λ₁–λ₄) that control color saturation, pattern density, compositional symmetry, and brushstroke density through a diagonal style matrix. The matrix conditions generation via cross-attention style tokens and FiLM/AdaIN feature modulation, while symbol-aware supervision improves cultural motif fidelity. Experiments on a 6,000-image expert-annotated Dunhuang dataset achieve 83.50% emotion-recognition accuracy, 78.90% emotion–style alignment, 82.20% cultural-symbol accuracy (micro-F1 × 100), and 0.68 s runtime at 512 × 512.
Zitieren
APA
Song, Y. (2026). Generation of Dunhuang Murals Using Emotion Vectors and Style Matrix Control. International Journal of Computational Intelligence Systems, 19(1). https://doi.org/10.1007/s44196-026-01346-4
Vancouver
Song Y. Generation of Dunhuang Murals Using Emotion Vectors and Style Matrix Control. International Journal of Computational Intelligence Systems. 2026;19(1). doi:10.1007/s44196-026-01346-4
BibTeX
@article{Song2026,
author = {Song, Yuchen},
title = {Generation of Dunhuang Murals Using Emotion Vectors and Style Matrix Control},
journal = {International Journal of Computational Intelligence Systems},
year = {2026},
volume = {19},
number = {1},
doi = {10.1007/s44196-026-01346-4}
}
EUVIMED