广师大多模态细粒度标签服装数据集

广师大多模态细粒度标签服装数据集在Locarno分类的基础上，采用人工标记的方式来构建。数据集包括背心，短裤，长裙，短袖，手套，袜子，胸罩，外套，长袖，短裙这10类外观设计专利图像。数据集包括30000个专利共60000张专利图像，分类为10个类别，每个类别包含6000张彩色外观专利图片，每张图像的空间分辨率有三种格式28×28、32×32、224×224像素，每个类5000张属于训练集，1000张属于测试集。需要指出的是每个专利提供了前后两个视图图像反映设计特征。与Fashion-MNIST、CIFAR-10数据集相比，本数据集图像具有外观专利的特点，图像质量好，图像背景清晰，特征明显。本数据集可以用于外观专利分类、检索等算法研究和应用开发。

Guangdong Normal University Multimodal Fine-grained Labeled Clothing Dataset is constructed based on the Locarno Classification system via manual annotation. The dataset includes 10 categories of appearance design patent images, specifically vest, shorts, long dress, short-sleeve top, gloves, socks, bra, coat, long-sleeve top, and short dress. In total, the dataset contains 30,000 patents corresponding to 60,000 patent images, which are categorized into the 10 aforementioned classes. Each class has 6,000 color appearance design patent images, with three spatial resolution specifications: 28×28, 32×32, and 224×224 pixels. For each category, 5,000 images are allocated to the training set, while the remaining 1,000 images are reserved for the test set. It is worth noting that each patent provides two view images (front and rear) to reflect its design characteristics. Compared with mainstream datasets such as Fashion-MNIST and CIFAR-10, the images in this dataset exhibit the unique features of appearance design patents, with high image quality, clear background, and salient distinct features. This dataset can be utilized for algorithm research and application development including appearance design patent classification and retrieval.

搜集汇总

背景与挑战

背景概述

广师大多模态细粒度标签服装数据集包含10类服装的60000张专利图像，每类6000张，图像分辨率有三种格式（28×28、32×32、224×224像素），适用于外观专利分类和检索研究。数据集具有清晰的图像背景和明显的特征，为算法研究和应用开发提供了高质量的数据支持。