Implementation of Text-to-Image Generators in the Development of the Usability Interface for the Construction of a Web Page

Chávez P.; Ticona W. · Proceedings of the 14th International Conference on Cloud Computing, Data Science and Engineering, Confluence 2024 (2024) · DOI: 10.1109/confluence60223.2024.10463436

Resumen

The generation of high-quality images with specific styles using Text-to-image Generators became more accessible due to the distribution of different diffusion models, which through the methodology Diffusion Training and Dreambooth technology. The main objective is getting a better efficiency in the development of a Usability Interface (UI) for a website. The members of the development refined Stable Diffusion models, to able to generate images with the styles: 'Facebook Alegria' (Alegria style) and 'black silhouette icons' (icon style). With the speed and simplicity of image generation, the team was able to reduce cost and times in the development. From the total number of images generated, 4 images with Alegria Style and 17 with icon style were chosen to add at the UI. The main result of the research showed that after the implementation, there was a significative change with respect to other work previously performed in the same company. The hours spent on frontend development were reduced by 81.65%, and and costs were reduced by 22.80%. In conclusion, the Post-test collected indicates notable improvements in the efficiency of the development. The implementation of text-to-imagen generations has proven its effectiveness in to reduce the cost and time of UI development.

Instituciones: Universidad ESAN; Universidad Tecnológica del Perú