arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2302.09795cs.LGcs.CVstat.ML
收藏

视觉表示中风格与内容的简单解耦

Simple Disentanglement of Style and Content in Visual Representations

Rensselaer Polytechnic Institute(伦斯勒理工学院);University of Michigan(密歇根大学);IBM Research(IBM研究院);MIT-IBM Watson AI Lab(麻省理工-IBM沃森人工智能实验室)

Lilian Ngweta, Subha Maity, Alex Gittens, Yuekai Sun, Mikhail Yurochkin

AI总结:

本文提出一种基于概率模型的简单后处理框架,从预训练视觉表示中解耦内容与风格,并在风格相关分布偏移下显著提升领域泛化性能。

摘要:

学习具有可解释特征的视觉表示,即解耦表示,仍然是一个具有挑战性的问题。现有方法取得了一定成功,但难以应用于像 ImageNet 这样的大规模视觉数据集。在这项工作中,我们提出了一个简单的后处理框架,用于从预训练视觉模型学习到的表示中解耦内容和风格。我们将预训练特征概率性地建模为潜在内容因子和风格因子的线性纠缠组合,并基于该概率模型开发了一种简单的解耦算法。我们证明了该方法在理论上能够解耦内容和风格特征,并通过实验验证了其有效性。当分布偏移由风格变化或与风格相关的虚假相关性引起时,我们后处理得到的特征能够显著提升领域泛化性能。

英文摘要:

Learning visual representations with interpretable features, i.e., disentangled representations, remains a challenging problem. Existing methods demonstrate some success but are hard to apply to large-scale vision datasets like ImageNet. In this work, we propose a simple post-processing framework to disentangle content and style in learned representations from pre-trained vision models. We model the pre-trained features probabilistically as linearly entangled combinations of the latent content and style factors and develop a simple disentanglement algorithm based on the probabilistic model. We show that the method provably disentangles content and style features and verify its efficacy empirically. Our post-processed features yield significant domain generalization performance improvements when the distribution shift occurs due to style changes or style-related spurious correlations.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑