完美颜色等变的超环面覆盖
A Hypertoroidal Covering for Perfect Color Equivariance
AI总结:
提出一种通过将区间值提升到圆上的双覆盖来构建真正等变的颜色等变架构,解决了先前方法中近似饱和度和亮度为1D平移带来的伪影问题,在细粒度分类和医学成像等任务上提升了性能。
摘要:
当输入图像的颜色分布在推理时发生变化,传统神经网络架构的性能会显著下降。一些研究者已开始将颜色几何的先验知识融入神经网络设计。这些颜色等变架构将色调变化建模为2D旋转,饱和度和亮度变换建模为1D平移。虽然这种方法在多种情况下提高了神经网络对颜色变化的鲁棒性,但我们发现将饱和度和亮度(区间值量)近似为1D平移会引入明显的伪影。本文提出了一种真正等变的颜色等变架构。我们不再用实直线近似区间,而是将区间上的值提升到圆上的值(双覆盖),并在其上构建等变表示。我们的方法解决了先前方法的近似伪影问题,提高了可解释性和泛化能力,并在细粒度分类和医学成像等任务上取得了优于传统和等变基线的预测性能。超越颜色范畴,我们提出的提升方法还可以扩展到尺度等几何变换。
英文摘要:
When the color distribution of input images changes at inference, the performance of conventional neural network architectures drops considerably. A few researchers have begun to incorporate prior knowledge of color geometry in neural network design. These color equivariant architectures have modeled hue variation with 2D rotations, and saturation and luminance transformations as 1D translations. While this approach improves neural network robustness to color variations in a number of contexts, we find that approximating saturation and luminance (interval valued quantities) as 1D translations introduces appreciable artifacts. In this paper, we introduce a color equivariant architecture that is truly equivariant. Instead of approximating the interval with the real line, we lift values on the interval to values on the circle (a double-cover) and build equivariant representations there. Our approach resolves the approximation artifacts of previous methods, improves interpretability and generalizability, and achieves better predictive performance than conventional and equivariant baselines on tasks such as fine-grained classification and medical imaging tasks. Going beyond the context of color, we show that our proposed lifting can also extend to geometric transformations such as scale.
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。