Davide Crociati
• Davide Sonno
abstract
Generative Text-to-Image (T2I) models have gained more and more popularity,
enabling users to easily produce detailed images from textual prompts. However,
these models often encode and reproduce the social or cultural biases present in their
training data. Such biases can influence people perception and reinforce stereotypes.
In this project we aim to explore and identify patterns of bias in six of the current
popular T2I models.
outcomes