Move Anything with Layered Scene Diffusion

arXiv:2404.07178v1 Announce Type: new
Abstract: Diffusion models generate images with an unprecedented level of quality, but how can we freely rearrange image layouts? Recent works generate controllable scenes via learning spatially disentangled latent codes, but these methods do not apply to diffusion models due to their fixed forward process. In this work, we propose SceneDiffusion to optimize a layered scene representation during the diffusion sampling process. Our key insight is that spatial disentanglement can be obtained by jointly denoising scene …

abstract apply arxiv cs.cv diffusion diffusion models generate image images process quality type via work