Unbiased Manifold Augmentation for Coarse Class Subdivision

Abstract

Coarse Class Subdivision (CCS) is important for many practical applications, where the training set originally annotated for a coarse class (e.g. bird) needs to further support its sub-classes recognition (e.g. swan, crow) with only very few fine-grained labeled samples. From the perspective of causal representation learning, these sub-classes inherit the same determinative factors of the coarse class, and their difference lies only in values. Therefore, to support the challenging CCS task with minimum fine-grained labeling cost, an ideal data augmentation method should generate abundant variants by manipulating these sub-class samples at the granularity of generating factors. For this goal, traditional data augmentation methods are far from sufficient. They often perform in highly-coupled image or feature space, thus can only simulate global geometric or photometric transformations. Leveraging the recent progress of factor-disentangled generators, Unbiased Manifold Augmentation (UMA) is proposed for CCS. With a controllable StyleGAN pre-trained for a coarse class, an approximate unbiased augmentation is conducted on the factor-disentangled manifolds for each sub-class, revealing the unbiased mutual information between the target sub-class and its determinative factors. Extensive experiments have shown that in the case of small data learning (less than 1% fine-grained samples of commonly used), our UMA can achieve 10.37% average improvement compared with existing data augmentation methods. On challenging tasks with severe bias, the accuracy is improved by up to 16.79%. We release our code at https://github.com/leo-gb/UMA.