Large Labelled Logo Dataset (L3D)

Large Labelled Logo Dataset (L3D) 是由巴塞罗那超级计算中心和马德里自治大学创建的多用途、手工标注的持续增长数据集。该数据集包含约77万张256x256像素的彩色RGB图像，这些图像从欧盟知识产权局（EUIPO）的公开注册中提取。每张图像都关联有多个标签，用于分类图像中的图形和文本元素。这些标注由EUIPO评估人员使用维也纳分类法进行分类，该分类法是一种图形标记的层次分类。L3D数据集的应用领域包括商标分类、商标生成、光学字符识别等，旨在解决商标图像的自动化处理和分析问题。

Large Labelled Logo Dataset (L3D) is a multi-purpose, manually annotated, continuously growing dataset created by the Barcelona Supercomputing Center and the Autonomous University of Madrid. This dataset contains approximately 770,000 256×256-pixel color RGB images extracted from public registrations of the European Union Intellectual Property Office (EUIPO). Each image is associated with multiple labels for classifying the graphical and textual elements within it. These annotations are classified by EUIPO examiners using the Vienna Classification, a hierarchical taxonomy for graphic marks. Application domains of the L3D dataset include trademark classification, trademark generation, optical character recognition (OCR) and other related fields, aiming to address the issues of automated processing and analysis of trademark images.