FashionEdit
收藏资源简介:
FashionEdit数据集是一个模拟真实世界服装设计工作流程的时尚数据集,用于评估基于图像到提示的细粒度定制服装设计。数据集由香港理工大学和中国石油大学(华东)的研究人员创建,包含从DeepFashion-MultiModal数据集生成的11546张图像及其与原始图像的差异,差异包括描述和从原始图像中裁剪的区域。数据集旨在解决当前生成式AI模型在服装设计定制化方面存在的文本描述模糊问题,通过结合文本提示和图像提示进行迭代图像编辑,以提高用户满意度和设计质量。
The FashionEdit Dataset is a fashion dataset that simulates real-world clothing design workflows, dedicated to evaluating image-to-prompt based fine-grained customized clothing design. It was created by researchers from The Hong Kong Polytechnic University and China University of Petroleum (East China). The dataset contains 11,546 images generated from the DeepFashion-MultiModal dataset, along with their differences from the original images, which include descriptive annotations and cropped regions extracted from the original images. This dataset aims to address the issue of vague text descriptions in current generative AI models for customized clothing design, and enables iterative image editing by combining both text prompts and image prompts to improve user satisfaction and design quality.

- 1通过香港理工大学,中国; 中国石油大学(华东),中国 · 2025年



