HyperAI

MSEmbGAN Multi-needle Embroidery Dataset

Date

8 months ago

Size

1000.37 MB

Organization

Publish URL

drive.google.com

* This dataset supports online use.Click here to jump.

This dataset is the research result of embroidery image generation by Hu Xinrong's research group from the School of Computer and Artificial Intelligence of Wuhan Textile University in 2024. The relevant paper is "MSEmbGAN: Multi-Stitch Embroidery Synthesis via Region-Aware Texture Generation", has been accepted by IEEE Transactions on Visualization and Computer Graphics (TVCG), with Professor Sheng Bin from the School of Computer Science and Engineering at Shanghai Jiao Tong University as the corresponding author. This method ensures that every stitch in the generated embroidery meets professional standards, supports designers in producing works that meet high aesthetic and technical standards, and significantly reduces the time spent on manual adjustments.

This dataset is the first embroidery dataset that is detailedly annotated with single-needle and multi-needle labels. More than 30K images, including embroidery images and corresponding content images, were produced by professional embroidery software (Wilcom 9.0). All images were resized to a resolution of 256 × 256. This paper contributes the constructed multi-needle embroidery dataset to other researchers in this research field. The dataset in this paper annotates one multi-needle type and three single-needle types, and the embroidery images are rendered by professional embroidery design software. The dataset contains more than 30,000 aligned or unaligned embroidery and content images. For the production of the multi-needle embroidery dataset, the steps for dataset image production are as follows:

  • Creating an embroidery dataset: Embroidery designers use professional embroidery software (Wilcom 9.0) to design and create embroidery patterns and render the corresponding embroidery images. These results are the embroidery age in the dataset and the ground truth images in this work. The following figure shows a small part of the dataset images established in this paper.
  • Drawing content images: Before making an embroidery board, the embroiderer must first draw a content image containing embroidery color information as a template. Most content images have simple colors and clear shapes, allowing faster network convergence.
  • Design of stitches: For content images with different shapes, a stitch must be selected to fill each region. The embroidery designer matches the shape of each region with an appropriate stitch type. In addition, the relevant parameters of each stitch (such as spacing and direction) must be reasonably set for subsequent embroidery rendering tasks.
MSEmb-dataset.torrent
Seeding 2Downloading 0Completed 147Total Downloads 386
  • MSEmb-dataset/
    • README.md
      2.68 KB
    • README.txt
      5.37 KB
      • data/
        • 多针刺绣数据集.zip
          1000.37 MB