Fashion-Intelligence/ST-Net-dataset

---
task_categories:
- image-to-image
size_categories:
- 10K<n<100K
---

This dataset was constructed for paper **Towards Intelligent Design: A Self-Driven Framework for Collocated Clothing Synthesis Leveraging Fashion Styles and Textures**.
You need to fill in the <a href="https://forms.gle/9M37JCGdkyLRKRxu9">form</a> to get password for unzipping files.

# Details

This dataset was constructed for self-supervised collocated clothing synthesis (CCS) task. It encompasses 30, 876 images, covering both upper and lower clothing items. We partition the dataset into training and test sets, maintaining a ratio of 4:1, for both 'upper &rarr; lower' and 'lower &rarr; upper' settings. Within this dataset, there exist instances where certain fashion items recur. To ensure non-overlap between the training and test sets, we employ a graph-based method and similarity comparison with learned perceptual image patch similarity (LPIPS). This ensures that the source domain images in any transformation direction do not intersect. For instance, in the transition direction of 'upper &rarr; lower', there is no intersection between the upper clothing images in the training set and those in the test set. All images in the dataset are with a resolution of 256 x 256 pixels.

# Citation
If you use this dataset in your work, please cite it as:
```
@inproceedings{dong2024towards,
title={Towards Intelligent Design: A Self-Driven Framework for Collocated Clothing Synthesis Leveraging Fashion Styles and Textures},
author={Dong, Minglong and Zhou, Dongliang and Ma, Jianghong and Zhang, Haijun},
booktitle={ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
pages={3725--3729},
year={2024},
organization={IEEE}
}
```

原始信息汇总

数据集概述

任务类别

• 图像到图像转换

大小类别

• 数据量介于10,000至100,000之间

数据集构建目的

• 用于论文《Towards Intelligent Design: A Self-Driven Framework for Collocated Clothing Synthesis Leveraging Fashion Styles and Textures》
• 针对自监督的同位服装合成（CCS）任务

数据集详情

• 包含30,876张图像，涵盖上衣和下衣
• 分为训练集和测试集，比例为4:1，适用于上衣到下衣和下衣到上衣两种设置
• 采用图方法和学习感知图像块相似度（LPIPS）确保训练集和测试集之间无重叠
• 所有图像分辨率为256 x 256像素

引用信息

• 若使用此数据集，请引用：

@inproceedings{dong2024towards, title={Towards Intelligent Design: A Self-Driven Framework for Collocated Clothing Synthesis Leveraging Fashion Styles and Textures}, author={Dong, Minglong and Zhou, Dongliang and Ma, Jianghong and Zhang, Haijun}, booktitle={ICASSP 2024-2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, pages={3725--3729}, year={2024}, organization={IEEE} }