Exploring and Utilizing Pattern Imbalance

In this paper, we identify pattern imbalance from several aspects, and further develop a new training scheme to avert pattern preference as well as spurious correlation. In contrast to prior methods which are mostly concerned with category or domain granularity, ignoring the potential finer structure that existed in datasets, we give a new definition of seed category as an appropriate optimization unit to distinguish different patterns in the same category or domain. Extensive experiments on domain generalization datasets of diverse scales demonstrate the effectiveness of the proposed method. * Corresponding author. al. [42] argue that there exist sub-networks with preferable domain generalization ability in the model and represent the sub-network through a learnable mask. Nam et al. [28] assume that the spurious features are generally embodied in the texture or style of the image. They design SagNet to decouple the content and style of the image, impelling the feature extractor to pay more attention to the content information. Most of the above methods manually design specific model structures to handle domain generalization.