Developing alternative text using neural networks for image generation: learning outcomes dataset

Published: 14 May 2023| Version 1 | DOI: 10.17632/c787ftvysg.1
Contributors:
Yekaterina Kosova,
,

Description

The dataset contains the results of training future web developers on how to develop alt text for images. The study was carried out in April-May 2023. The Kandinsky 2.1 neural network (https://fusionbrain.ai/diffusion) was used to evaluate the image quality. 17 students prepared alt texts for three photos. The results were evaluated before and after using the neural network. The data set is presented in a Microsoft Excel table on the Alt_Text_Result sheet with the following fields: record number; participant number; photo number; photo alt text; number of characters in alt text; quality assessment of alt text; links to generated images; degree of similarity between generated images and original photos. The original photos are placed on the Alt_Text_Photos sheet. Alternative descriptions are presented in Russian.

Files

Categories

World Wide Web, Text Editing, e-Learning, Accessibility Issue, Image Analysis, Neural Network

Licence