Domain-aware Visual Bias Eliminating for Generalized Zero-Shot Learning
This addresses the challenge of domain bias in zero-shot learning, which is incremental as it builds on prior work by focusing on semantic-free representations.
The paper tackles the biased recognition problem in generalized zero-shot learning by proposing a Domain-aware Visual Bias Eliminating (DVBE) network that uses complementary semantic-free and semantic-aligned visual representations, resulting in an average 5.7% improvement over existing methods on five benchmarks.
Recent methods focus on learning a unified semantic-aligned visual representation to transfer knowledge between two domains, while ignoring the effect of semantic-free visual representation in alleviating the biased recognition problem. In this paper, we propose a novel Domain-aware Visual Bias Eliminating (DVBE) network that constructs two complementary visual representations, i.e., semantic-free and semantic-aligned, to treat seen and unseen domains separately. Specifically, we explore cross-attentive second-order visual statistics to compact the semantic-free representation, and design an adaptive margin Softmax to maximize inter-class divergences. Thus, the semantic-free representation becomes discriminative enough to not only predict seen class accurately but also filter out unseen images, i.e., domain detection, based on the predicted class entropy. For unseen images, we automatically search an optimal semantic-visual alignment architecture, rather than manual designs, to predict unseen classes. With accurate domain detection, the biased recognition problem towards the seen domain is significantly reduced. Experiments on five benchmarks for classification and segmentation show that DVBE outperforms existing methods by averaged 5.7% improvement.