Generative Classifiers as a Basis for Trustworthy Image Classification
Generative Classifiers as a Basis for Trustworthy Image Classification
复制标题
生成分类器作为可信图像分类的基础
DOI：
10.1109/cvpr46437.2021.00299
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
C. Rother
中科院分区：
文献类型：
--
作者：
Radek Mackowiak;Lynton Ardizzone;Ullrich Kothe;C. Rother
随着深度学习系统的成熟，可信性在模型评估中的重要性日益凸显。我们将可信性理解为可解释性和健壮性的结合。生成分类器(GC)是一类很有前途的模型，据说可以自然地实现这些特性。然而，这一点过去大多是在MNIST和CIFAR等简单数据集上演示的。在这项工作中，我们首先开发了一个体系结构和培训方案，允许GC在与实际计算机视觉更相关的复杂性水平上操作，即ImageNet挑战。其次，我们展示了GC在可信图像分类方面的巨大潜力。与前馈模型相比，可解释性和稳健性的某些方面都得到了极大的改进，即使GC只是在幼稚的情况下应用。虽然并不是所有的可信度问题都被完全解决，但我们观察到GC是进一步算法和修改的一个非常有希望的基础。我们发布我们训练过的模型以供下载，希望它可以作为其他生成性分类任务的起点，就像预先训练的ResNet体系结构用于区分分类一样。代码：githeb.com/vll-hd/trustworth_gcs
With the maturing of deep learning systems, trustworthiness is becoming increasingly important for model assessment. We understand trustworthiness as the combination of explainability and robustness. Generative classifiers (GCs) are a promising class of models that are said to naturally accomplish these qualities. However, this has mostly been demonstrated on simple datasets such as MNIST and CIFAR in the past. In this work, we firstly develop an architecture and training scheme that allows GCs to operate on a more relevant level of complexity for practical computer vision, namely the ImageNet challenge. Secondly, we demonstrate the immense potential of GCs for trustworthy image classification. Explainability and some aspects of robustness are vastly improved compared to feed-forward models, even when the GCs are just applied naively. While not all trustworthiness problems are solved completely, we observe that GCs are a highly promising basis for further algorithms and modifications. We release our trained model for download in the hope that it serves as a starting point for other generative classification tasks, in much the same way as pretrained ResNet architectures do for discriminative classification.Code: github.com/VLL-HD/trustworthy_GCs
DOI：
--
发表时间：
2020-06
期刊：
ArXiv
影响因子：
--
作者：
P. Kirichenko;Pavel Izmailov;A. Wilson
通讯作者：
P. Kirichenko;Pavel Izmailov;A. Wilson