纺织科技进展

基于Stable Diffusion和LoRA模型的数据采集、特征提取和模型训练，构建AIGC范式用于生成具有巫溪嫁花风格的纹样。研究创新点在于将LoRA模型应用于巫溪嫁花纹样的智能化设计，实现对传统纹样的深度学习和风格迁移；同时采用SSIM和FID指标客观评估生成图像的质量，确保设计的准确性和创新性。结果显示，构建的AIGC范式能有效生成具有巫溪嫁花风格的纹样，为传统纹样的数字化传承与创新提供了新路径，后续研究将致力于优化模型的泛化能力，以期为传统艺术的数字转化及应用提供坚实的技术支持。

Through data collection, feature extraction, and model training based on the Stable Diffusion and LoRA models, an AIGC paradigm was constructed to generate patterns with the style of Wuxi Jiahua. The innovation lies in the application of the LoRA model to the intelligent design of Wuxi Jiahua patterns, achieving deep learning and style transfer of traditional patterns. At the same time, SSIM and FID were employed to objectively evaluate the quality of generated images, ensuring accuracy and innovation in the design. Experimental results demonstrated that the constructed AIGC paradigm could effectively generate patterns with Wuxi Jiahua characteristics, providing a new path for digital inheritance and innovation of traditional patterns. Further research would focus on optimizing the model′s generalization ability to offer technical support for digital transformation and application of traditional art.