Scholarlink

<p>Traditional Miao embroidery features intricate pattern structures. The hand embroidery restoration techniques are extremely labor-intensive and time-consuming. In order to improve the efficiency of embroidery image restoration, this paper develops a framework for Miao embroidery pattern image restoration. The framework combines generative adversarial network with U-Net. The U-Net incorporates gated convolutions and spatial-channel attention mechanisms to enhance the model’s ability to learn and reconstruct the intricate textures and structures of the embroidery. The proposed algorithm is compared with the current mainstream algorithms using PSNR, SSIM, and LPIPS metrics. The results show that this algorithm performs better than other methods in Miao embroidery restoration. The source code and datasets used in this study are available at Zenodo (DOI: 10.5281/zenodo.12759273).</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Restoring intricate Miao embroidery patterns: a GAN-based U-Net with spatial-channel attention

  • 期刊论文

  • 发表日期：

    2025年2月14日

查看全文
  • Cheng Zhong,
  • Xiaomin Yu,
  • Huan Xia,
  • Rongdong Xie,
  • Qingyi Xu

摘要

Traditional Miao embroidery features intricate pattern structures. The hand embroidery restoration techniques are extremely labor-intensive and time-consuming. In order to improve the efficiency of embroidery image restoration, this paper develops a framework for Miao embroidery pattern image restoration. The framework combines generative adversarial network with U-Net. The U-Net incorporates gated convolutions and spatial-channel attention mechanisms to enhance the model’s ability to learn and reconstruct the intricate textures and structures of the embroidery. The proposed algorithm is compared with the current mainstream algorithms using PSNR, SSIM, and LPIPS metrics. The results show that this algorithm performs better than other methods in Miao embroidery restoration. The source code and datasets used in this study are available at Zenodo (DOI: 10.5281/zenodo.12759273).