20万组多国地标建筑图文数据

20万组多国地标建筑图文数据。每组数据包含一张图片+中英文描述文档。国内地标包括8万组，其他国家地标12万组。国家分布不少于20个，包括美国、英国、法国、俄罗斯、德国等国家。每个地标对应的图片张数分布在1-10张，包括不同角度、不同距离、不同时间段的地标信息。地标类型包括多种类型的地标建筑，比如商业大厦、古老建筑、纪念碑、图书馆、景点等。在标注内容方面，对地标国家、地标城市、地标具体地点、地标分类、地标描述进行标注。数据可用于地标建筑识别和分析。

200,000 sets of multi-national landmark building image-text paired data are collected. Each set consists of one image and a bilingual (Chinese and English) descriptive document. Specifically, there are 80,000 sets of domestic landmarks within China and 120,000 sets of landmarks from other countries. The dataset covers no fewer than 20 countries, including the United States, the United Kingdom, France, Russia, Germany and others. Each landmark is paired with 1 to 10 images, which capture the landmark from different angles, shooting distances and time periods. The landmark categories cover various types, such as commercial skyscrapers, ancient architectures, monuments, libraries, tourist attractions and more. For annotation, the dataset provides labels for the landmark's country, city, specific location, category and descriptive information. This dataset can be utilized for landmark recognition and analysis tasks.

搜集汇总

背景与挑战

背景概述

该数据集包含20万组多国地标建筑的图文数据，覆盖国内外多种地标类型，标注详细且准确率高，适用于地标建筑识别和分析任务。