Stable Diffusion is a text-to-image generation model that is based on latent diffusion. It works by first translating the textual prompt into a multidimensional latent space, which can be seen as an internal representation of a conceptual space. For other kinds of generative models, it has been argued that relationships between concepts can be deduced from the geometrical properties of the latent space. In this paper we explore this claim for a pre-trained Stable Diffusion model. In particular, we verify its capabilities to produce images that blend two concepts without any fine-tuning.
Melzi, S., Penaloza, R., Raganato, A. (2023). Does Stable Diffusion Dream of Electric Sheep?. In Proceedings of The Seventh Image Schema Day co-located with The 20th International Conference on Principles of Knowledge Representation and Reasoning (KR 2023) (pp.1-11). CEUR-WS.
Does Stable Diffusion Dream of Electric Sheep?
Melzi S.;Penaloza R.
;Raganato A.
2023
Abstract
Stable Diffusion is a text-to-image generation model that is based on latent diffusion. It works by first translating the textual prompt into a multidimensional latent space, which can be seen as an internal representation of a conceptual space. For other kinds of generative models, it has been argued that relationships between concepts can be deduced from the geometrical properties of the latent space. In this paper we explore this claim for a pre-trained Stable Diffusion model. In particular, we verify its capabilities to produce images that blend two concepts without any fine-tuning.
| File | Dimensione | Formato | |
|---|---|---|---|
|
Melzi-2023-ISD-AAM.pdf
accesso aperto
Tipologia di allegato:
Author’s Accepted Manuscript, AAM (Post-print)
Licenza:
Creative Commons
Dimensione
3.46 MB
Formato
Adobe PDF
|
3.46 MB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
