Self-supervised learning (SSL) as an effective paradigm of representation learning has achieved tremendous success on various curated datasets in diverse scenarios. Nevertheless, when facing the long-tailed distribution in real-world applications, it is still hard for existing methods to capture transferable and robust representation. Conventional SSL methods, pursuing sample-level uniformity, easily leads to representation learning disparity where head classes dominate the feature regime but tail classes passively collapse. To address this problem, we propose a novel Geometric Harmonization (GH) method to encourage category-level uniformity in representation learning, which is more benign to the minority and almost does not hurt the majority under long-tailed distribution. Specially, GH measures the population statistics of the embedding space on top of self-supervised learning, and then infer an fine-grained instance-wise calibration to constrain the space expansion of head classes and avoid the passive collapse of tail classes. Our proposal does not alter the setting of SSL and can be easily integrated into existing methods in a low-cost manner. Extensive results on a range of benchmark datasets show the effectiveness of GH with high tolerance to the distribution skewness. Our code is available at https://github.com/MediaBrain-SJTU/Geometric-Harmonization.

翻译：自监督学习作为表示学习的有效范式，已在多种场景下的各类精心整理数据集中取得了巨大成功。然而，在面对现实应用中存在的长尾分布时，现有方法仍难以捕获可迁移且鲁棒的表示。传统自监督学习方法追求样本级均匀性，容易导致表示学习差异，即头部类别主导特征空间而尾部类别被动坍缩。为解决这一问题，我们提出了一种新颖的几何调和法，通过鼓励表示学习中的类别级均匀性，在长尾分布下对少数类更友好且几乎不影响多数类。具体而言，GH在自监督学习的基础上度量嵌入空间的群体统计量，进而推断出细粒度的实例级校正，以约束头部类别的空间扩张并避免尾部类别的被动坍缩。我们的方法不改变自监督学习的设置，且能以低成本方式轻松集成到现有方法中。在多个基准数据集上的广泛结果显示了GH对分布偏斜的高容忍度及其有效性。我们的代码已开源在 https://github.com/MediaBrain-SJTU/Geometric-Harmonization。