Mutual Balancing in State-Object Components for Compositional Zero-Shot Learning
This addresses a key bottleneck in CZSL for recognizing unseen compositions, offering a domain-specific improvement that enhances various frameworks.
The paper tackles the imbalance in visual deviation between semantic labels and actual features in Compositional Zero-Shot Learning (CZSL), proposing a method that splits classification into two processes to analyze component entanglement and modify training, resulting in significant performance improvements over state-of-the-art on benchmarks like MIT-States, UT-Zappos, and C-GQA.
Compositional Zero-Shot Learning (CZSL) aims to recognize unseen compositions from seen states and objects. The disparity between the manually labeled semantic information and its actual visual features causes a significant imbalance of visual deviation in the distribution of various object classes and state classes, which is ignored by existing methods. To ameliorate these issues, we consider the CZSL task as an unbalanced multi-label classification task and propose a novel method called MUtual balancing in STate-object components (MUST) for CZSL, which provides a balancing inductive bias for the model. In particular, we split the classification of the composition classes into two consecutive processes to analyze the entanglement of the two components to get additional knowledge in advance, which reflects the degree of visual deviation between the two components. We use the knowledge gained to modify the model's training process in order to generate more distinct class borders for classes with significant visual deviations. Extensive experiments demonstrate that our approach significantly outperforms the state-of-the-art on MIT-States, UT-Zappos, and C-GQA when combined with the basic CZSL frameworks, and it can improve various CZSL frameworks. Our codes are available on https://anonymous.4open.science/r/MUST_CGE/.