Learning Cross-domain Semantic-Visual Relationships for Transductive Zero-Shot Learning
This work addresses domain adaptation challenges in zero-shot learning for recognizing new classes, representing an incremental improvement over existing methods.
The paper tackles the domain discrepancy problem in transductive zero-shot learning by proposing the Transferrable Semantic-Visual Relation (TSVR) approach, which redefines image recognition as predicting similarity labels for semantic-visual fusions and uses Domain-Specific Batch Normalization to align domains, achieving unspecified performance gains.
Zero-Shot Learning (ZSL) learns models for recognizing new classes. One of the main challenges in ZSL is the domain discrepancy caused by the category inconsistency between training and testing data. Domain adaptation is the most intuitive way to address this challenge. However, existing domain adaptation techniques cannot be directly applied into ZSL due to the disjoint label space between source and target domains. This work proposes the Transferrable Semantic-Visual Relation (TSVR) approach towards transductive ZSL. TSVR redefines image recognition as predicting the similarity/dissimilarity labels for semantic-visual fusions consisting of class attributes and visual features. After the above transformation, the source and target domains can have the same label space, which hence enables to quantify domain discrepancy. For the redefined problem, the number of similar semantic-visual pairs is significantly smaller than that of dissimilar ones. To this end, we further propose to use Domain-Specific Batch Normalization to align the domain discrepancy.