Research on Traditional Costume Culture Information Extraction and Digital Reconstruction Methods Based on Artificial Intelligence

As a carrier of national cultural characteristics, traditional national costumes have an important position in national traditional culture. The research is based on artificial intelligence technology and analyzes the digital exploration of traditional dress culture. In view of the problem of poor accuracy rate of traditional dress pattern segmentation, after proposing a RSKP-UNet model integrating U-Net model and attention mechanism for traditional dress culture information extraction research, and introducing an IoU-based loss function, the model is trained and validated using self-built dataset. The LoRA fine-tuning technique is then combined with the stable diffusion model, and a discriminative network is introduced to establish the traditional dress pattern generation method with the improved stable diffusion model. The RSKP-UNet model has higher information extraction accuracy than the UNet model for different parts of the traditional dress, with the AP values ranging from 71.54% to 95.62%, which realizes effective extraction of the features of the traditional dress elements in the images. The test of the improved stable diffusion model revealed the good fit and image generation effects of the model, which demonstrate better experimental results in both quantitative and qualitative experiments. The study helps promote the preservation and innovative development of traditional dress culture.