Superpixel-based Visual Feature Enhancement for Compositional Zero-Shot Learning

doi:10.1016/j.ipm.2025.104414

Superpixel-based Visual Feature Enhancement for Compositional Zero-Shot Learning

2025-10-03

0

PRE

AI

DOI：10.1016/j.ipm.2025.104414

Original

Original request for help

Save

Abstract

En 中文

• A superpixel-based algorithm is designed to enable a finer granularity for feature extraction. • A Fourier spectral layer is introduced to capture global visual features from the frequency domain. • A long-range fusion module is proposed to recognize complex compositional relationships. • Experimental results demonstrate the superiority of the proposed approach over other CZSL methods.

Superpixel-based Visual Feature Enhancement for Compositional Zero-Shot Learning

Abstract

Journal

Organization