arXiv 2306.16934

DreamDiffusion: Generating High-Quality Images from Brain EEG Signals

By Yunpeng Bai, Xintao Wang, et al.

Published 2023-06-29

Mindmap

Browse the paper's core ideas, clusters, and relationships in a structured outline.

This paper introduces DreamDiffusion, a novel method for generating high-quality images directly from brain electroencephalogram (EEG) signals, without the need to translate thoughts into text. DreamDiffusion leverages pre-trained text-to-image models and employs temporal masked signal modeling to pre-train the EEG encoder for effective and robust EEG representations. Additionally, the method further leverages the C…

View the original paper on arXiv