Break-A-Scene: Extracting Multiple Concepts from a Single Image

Omri Avrahami, Kfir Aberman, Ohad Fried, Daniel Cohen-Or, Dani Lischinski

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Text-to-image model personalization aims to introduce a user-provided concept to the model, allowing its synthesis in diverse contexts. However, current methods primarily focus on the case of learning a single concept from multiple images with variations in backgrounds and poses, and struggle when adapted to a different scenario. In this work, we introduce the task of textual scene decomposition: given a single image of a scene that may contain several concepts, we aim to extract a distinct text token for each concept, enabling fine-grained control over the generated scenes. To this end, we propose augmenting the input image with masks that indicate the presence of target concepts. These masks can be provided by the user or generated automatically by a pre-trained segmentation model. We then present a novel two-phase customization process that optimizes a set of dedicated textual embeddings (handles), as well as the model weights, striking a delicate balance between accurately capturing the concepts and avoiding overfitting. We employ a masked diffusion loss to enable handles to generate their assigned concepts, complemented by a novel loss on cross-attention maps to prevent entanglement. We also introduce union-sampling, a training strategy aimed to improve the ability of combining multiple concepts in generated images. We use several automatic metrics to quantitatively compare our method against several baselines, and further affirm the results using a user study. Finally, we showcase several applications of our method.

Original languageEnglish
Title of host publicationProceedings - SIGGRAPH Asia 2023 Conference Papers, SA 2023
EditorsStephen N. Spencer
PublisherAssociation for Computing Machinery, Inc
ISBN (Electronic)9798400703157
DOIs
StatePublished - 10 Dec 2023
Event2023 SIGGRAPH Asia 2023 Conference Papers, SA 2023 - Sydney, Australia
Duration: 12 Dec 202315 Dec 2023

Publication series

NameProceedings - SIGGRAPH Asia 2023 Conference Papers, SA 2023

Conference

Conference2023 SIGGRAPH Asia 2023 Conference Papers, SA 2023
Country/TerritoryAustralia
CitySydney
Period12/12/2315/12/23

Keywords

  • multiple concept extraction
  • personalization
  • textual inversion

Fingerprint

Dive into the research topics of 'Break-A-Scene: Extracting Multiple Concepts from a Single Image'. Together they form a unique fingerprint.

Cite this