Nexdata/Landmark_Image_Dataset_200K_Global_Building_Photos_with_Captions

Description

200,000 Sets of Multi-country Landmark Buildings Image Caption Data. Each group of data consists of one images and Chinese-English description document. There are 80,000 sets of domestic landmarks and 120,000 sets of landmarks from other countries. The countries covered are at least 20, including the United States, the United Kingdom, France, Russia, Germany, etc. The number of images corresponding to each landmark ranges from 1 to 10, including information about the landmark from different angles, distances and time periods. The types of landmarks include various types of landmark buildings, such as commercial buildings, ancient buildings, monuments, libraries, scenic spots, etc. In terms of annotation content, the landmark's country, landmark city, specific location of the landmark, landmark category, and landmark description are all annotated. The data can be used for landmark building recognition and analysis.

For more details, please refer to the link: https://www.nexdata.ai/datasets/llm/1858?source=Huggingface

Specifications

Data size：

200,000 sets, each group of data contains one image and Chinese-English description document

Data Distribution：

80,000 sets of domestic landmarks, 120,000 sets of landmarks in other countries

Country Distribution：

not less than 20 countries are included, such as the United States, the United Kingdom, France, Russia, Germany, etc.

Landmark Distribution：

the number of images corresponding to each landmark ranges from 1 to 10, including information about the landmark from different angles, distances and time periods

Landmark Types：

it includes a variety of types of landmark buildings, such as commercial buildings, ancient buildings, monuments, libraries, attractions, etc.

Data Diversity：

different countries, different landmarks, different types

Resolution：

the resolution is no less than 1080p

Data Format：

the image format is .jpg or other common image formats, and the annotation document format is .json

Annotation Content：

the country of the landmark, the city of the landmark, the specific location of the landmark, landmark classification, landmark description were annotated

Accuracy rate：

the accuracy rate of landmark marking information shall not be lower than 97%. The landmark description text is objective and clear, with no obvious sentence errors, no sensitive content descriptions, and the sentence accuracy rate is no less than 97%

Licensing Information

Commercial License