Instance-aware visual-semantic interaction for zero-shot learning

doi:10.1016/j.eswa.2026.131682

Instance-aware visual-semantic interaction for zero-shot learning

2026-02-13

0

PRE

AI

DOI：10.1016/j.eswa.2026.131682

Original

Original request for help

Save

Abstract

En 中文

• We propose IVIS, a framework providing fine-grained, instance-level supervision. • IVIS performs instance-level semantic alignment to narrow the visual-semantic gap. • Learnable guidance tokens enable adaptive channel-wise fusion under global alignment.

Keywords:

IVIS

instance-level supervision

visual-semantic alignment

zero-shot learning

learnable guidance tokens

Instance-aware visual-semantic interaction for zero-shot learning

Abstract

Journal

Organization