Document is current

Article History

Received: 2 August 2023

Accepted: 4 February 2025

First Online: 13 March 2025

Declarations

:

: The authors declare that they have no Conflict of interest.

: As the use of machine learning in everyday life grows, it is relevant to consider the potential social impact of our work. Our work has the potential to be used for deep fake. Since our model can generate high-fidelity images with specific attributes, this even makes deep fake more flexible. On the other hand, our attribute-centric generative model is less affected by overrepresented attribute compositions in the dataset and can generate the images that match the given text. Therefore, our work contributes to the elimination of bias in generative models.