Multi-granularity contrastive zero-shot learning model based on attribute decomposition

Yuanlong Wang, Jing Wang, Yue Fan, Qinghua Chai, Hu Zhang, Xiaoli Li, Ru Li

Abstract: Highlights•Construct a Multi-granularity Contrastive Zero-shot Learning Model Based on Attribute Decomposition.•Introduce the global–local paradigm into the zero-shot image classification task, proposing a multi-granularity contrastive learning method.•Propose the attribute decomposition module, which decomposes attributes into key and common attributes to increase understanding of the role of different attributes.