Data set

The .csv generated by the image search provides the core naming data (images, responses, response counts) in long format. The full datasets (including target coordinates etc.) can be downloaded in .json and .tsv (tab-separated values) versions with the links below. In the .tsv version, nested data is formatted for use in Python. The GitHub repository contains scripts (in Python and R) illustrating access to the full data sets.

• English:
• manynames-en.tsv
• manynames-en.json
• Mandarin Chinese:
• manynames-zh.tsv
• manynames-zh.json

Images

The ManyNames images were sampled from the VisualGenome dataset. They were rescaled and displayed with a red bounding box for the data collection. This version of the images can be downloaded at manynames.upf.edu.

Scripts

The following scripts can be used to download a specific image set generated by the image search. To do so place the scripts in the same folder as the ManyNames_data.csv file generated with the image search and run the scripts.

• download image set (Python)
• download image set (R)

The following script can be used to recreate the ManyNames images directly from the VisualGenome originals. The scripts rescales the VisualGenome images and adds a red bounding box around the ManyNames target object.

• create images (Python)