CSCNET: Class-Specified Cascaded Network for Compositional Zero-Shot Learning

TL;DR

CSCNet introduces a cascaded framework for compositional zero-shot learning that improves attribute-object disentanglement by classifying primitives sequentially, leading to better recognition of novel compositions.

Contribution

The paper proposes CSCNet, a novel cascaded network that enhances A-O disentanglement by classifying primitives sequentially with a parametric classifier, advancing CZSL performance.

Findings

Achieves superior results compared to previous methods.

Effectively models contextual dependency between attribute and object primitives.

Improves matching between visual and semantic embeddings.

Abstract

Attribute and object (A-O) disentanglement is a fundamental and critical problem for Compositional Zero-shot Learning (CZSL), whose aim is to recognize novel A-O compositions based on foregone knowledge. Existing methods based on disentangled representation learning lose sight of the contextual dependency between the A-O primitive pairs. Inspired by this, we propose a novel A-O disentangled framework for CZSL, namely Class-specified Cascaded Network (CSCNet). The key insight is to firstly classify one primitive and then specifies the predicted class as a priori for guiding another primitive recognition in a cascaded fashion. To this end, CSCNet constructs Attribute-to-Object and Object-to-Attribute cascaded branches, in addition to a composition branch modeling the two primitives as a whole. Notably, we devise a parametric classifier (ParamCls) to improve the matching between visual and semantic embeddings. By improving the A-O disentanglement, our framework achieves superior results than previous competitive methods.