Drawing a portrait or imagining what anyone will look like twenty years from now is made possible by generative artificial intelligence (AI). Tools like Stable Diffusion, a machine learning model developed by Stability AI, is one of the applications capable of generating high-quality digital images from natural language descriptions. Just enter a short description and the machine will do the rest.
We can give it a description or read our thoughts, because Stable Diffusion is capable of reconstructing visual experiences from the activity of the human brain, according to a investigation published in biorxiv.
read also
“We show that our simple framework can reconstruct high-resolution images from brain activity with high semantic fidelity, without the need for complex deep generative model training or tuning,” explain Yu Takagi and Shinji Nishimoto, researchers at the Graduate School of Frontier. Biosciences from Osaka University (Japan).
with AI
Reconstruct visual experiences
The key was in the processes applied to transform brain activity into data, so that in this way they could be read and used by this tool.
The authors started from the idea that the reconstruction of visual experiences from the activity of the human brain “offers a unique way to understand how the brain represents the world and to interpret the connection between computer vision models and our visual system” .
Previous research had already delved into this question, but using overly complex systems: “Although deep generative models have recently been used for this task, the reconstruction of realistic images with high semantic fidelity remains a challenging problem. Here, we propose a new method based on a diffusion model (DM) to reconstruct images of human brain activity obtained through functional magnetic resonance imaging (fMRI).”
Takagi and Nishimoto reconstructed visual images from functional fMRI signals, then using the Stable Diffusion tool. The key was in the processes applied to transform brain activity into data, so that in this way they could be read and used by this tool.
“We quantitatively interpret each component of a latent diffusion model (LDM) from a neuroscientific perspective, assigning specific components to brain regions. We also present an objective interpretation of how the text-to-image conversion process implemented by an LDM incorporates the semantic information expressed by the conditional text, while at the same time maintaining the appearance of the original image.
read also
The authors acknowledge that previous studies had already managed to generate high-resolution images, but with more complex systems. They assure that this is the merit of their research and of Stable Diffusion. “Overall, our study proposes a promising method for reconstructing images of human brain activity and provides a new framework for understanding MD,” they argue.
Learning
Stable Diffusion is capable of drawing almost everything
Image generated by artificial intelligence, with Stable Diffusion
Stable Diffusion is a machine learning model developed by Stability AI prepared to generate high-quality digital images from natural language descriptions. Its popularity is increasing among users because it is easy to use, its results are accurate, and because it is not only capable of creating snapshots, but can edit or enhance them with new elements, depending on the instructions it receives from the user.
This AI tool is not without controversy. Some agencies or artists have taken legal action on the understanding that it infringes intellectual property rights, including copyright, by copying and processing millions of images posted on the internet.
#create #generative #artificial #intelligence #capable #drawing #thinking
