Crowdsourcing Thumbnail Captions via Time-Constrained Methods

Carlos Aguirre, Amama Mahmood and Chien-Ming Huang

Data

For more information in how we collected this data please see paper

Image Descritions

Image descriptions are stored in descriptions.csv. The file has the following columns:

• image_name - the ID of the image that matches MSCOCO image-ID
• condition - the name of the interface used to collect this description
• description - the text of the description
• time - amount of time in miliseconds (ms) that the worker took to write the description
• id - description ID, unique in this file

Human Ratings of Descriptions

A subset of the descriptions in descriptions.csv was rated based on 3 metrics: grammar, correctness, detail. The file contains the following columns:

• description_id - matches to id in descriptions.csv
• image_name - the ID of the image that matches MSCOCO image-ID
• fluency - (0-100) fluency of the description language
• correctness - (0-100) correctness of the description details based on image
• detail - (0-100) amount of detail contained in the description

Citing this Dataset

When referencing this dataset in your own manuscripts and publications, please use the following full citations:

[1] Aguirre, Carlos A., Amama Mahmood, and Chien-Ming Huang. "Crowdsourcing Thumbnail Captions via Time-Constrained Methods." 27th International Conference on Intelligent User Interfaces. 2022.