Separating anything from image in context

doi:10.1016/j.patcog.2026.113839

Separating anything from image in context

2026-04-24

0

PRE

AI

DOI：10.1016/j.patcog.2026.113839

Original

Original request for help

Save

Abstract

En 中文

• A generalist framework to separate arbitrary visual components. • A semantic in-context decoupling paradigm for open-world generalization. • Comprehensive experiments verify strong superiority across tasks.

Keywords:

visual separation

in-context learning

semantic decoupling

open-world generalization

image understanding

Separating anything from image in context

Abstract

Journal

Organization