FashionAI—服饰关键点定位数据集

为了推动AI技术与时尚行业的结合， 阿里巴巴“图像和美”团队联合香港理工大学纺织与制衣系联合推出业界首个同时符合服饰专业性和机器学习要求的大规模高质量时尚数据集FashionAI数据集。

服饰关键点定位数据集来自天池“FashionAI全球挑战赛——服饰关键点定位”竞赛：https://tianchi.aliyun.com/competition/entrance/231648/introduction 。研究人员可基于此数据集开展服饰关键点定位技术的研究，推动服饰部位对齐、服饰局部属性识别、服饰图像自动编辑等应用的效果提升。

To advance the integration of artificial intelligence (AI) technology and the fashion industry, Alibaba's "Image & Beauty" team, in collaboration with the Department of Textiles and Clothing of The Hong Kong Polytechnic University, has launched FashionAI, the industry's first large-scale high-quality fashion dataset that meets both professional apparel standards and machine learning requirements.

The clothing landmark localization dataset is derived from the Tianchi "FashionAI Global Challenge - Clothing Landmark Localization" competition: https://tianchi.aliyun.com/competition/entrance/231648/introduction. Researchers can utilize this dataset to conduct studies on clothing landmark localization technologies, so as to enhance the performance of applications including clothing part alignment, local clothing attribute recognition, and automatic clothing image editing.

搜集汇总

背景与挑战

背景概述

FashionAI—服饰关键点定位数据集是阿里巴巴与香港理工大学联合推出的业界首个大规模高质量时尚数据集，专注于女装五类服饰（上衣、外套、裤子、半身裙、连身裙）的关键点定位，包含约5.1GB训练数据和3.71GB测试数据。数据集采用专业标注，定义24个关键点并以三元组格式记录位置和可见性，旨在推动服饰部位对齐、属性识别等AI应用研究，源自天池竞赛并支持标准化评测。