README.md

July 12, 2022 · View on GitHub

Unbiased Manifold Augmentation for Coarse Class Subdivision

By Baoming Yan, KE GAO, Bo Gao, Lin Wang, Jiang Yang, Xiaobo Li.

This repo is the official implementation of "Unbiased Manifold Augmentation for Coarse Class Subdivision"

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%.

Getting Started

Progressive Sample Synthesis (PSS)

cd ./encoder4editing/scripts
bash inference.sh

Progressive Robust Learning (PRL)

cd ./ccs_training/
bash train.sh

Citing

...