Today's state-of-the-art machine vision models are vulnerable to image corruptions like blurring or compression artefacts, limiting their performance in many real-world applications. We here argue that popular benchmarks to measure model robustness against common corruptions (like ImageNet-C) underestimate model robustness in many (but not all) application scenarios. The key insight is that in many scenarios, multiple unlabeled examples of the corruptions are available and can be used for unsupervised online adaptation. Replacing the activation statistics estimated by batch normalization on the training set with the statistics of the corrupted images consistently improves the robustness across 25 different popular computer vision models. Using the corrected statistics, ResNet-50 reaches 62.2% mCE on ImageNet-C compared to 76.7% without adaptation. With the more robust DeepAugment+AugMix model, we improve the state of the art achieved by a ResNet50 model up to date from 53.6% mCE to 45.4% mCE. Even adapting to a single sample improves robustness for the ResNet-50 and AugMix models, and 32 samples are sufficient to improve the current state of the art for a ResNet-50 architecture. We argue that results with adapted statistics should be included whenever reporting scores in corruption benchmarks and other out-of-distribution generalization settings.

中文翻译：

---

通过协变量转移适应提高对常见腐败的鲁棒性

当今最先进的机器视觉模型容易受到模糊或压缩伪影等图像损坏的影响，从而限制了它们在许多实际应用中的性能。我们在这里认为，在许多（但不是全部）应用场景中，用于衡量模型针对常见损坏（如 ImageNet-C）稳健性的流行基准会低估模型的稳健性。关键的见解是，在许多场景中，可以使用多个未标记的损坏示例，可用于无监督的在线适应。用损坏图像的统计数据替换训练集上批量归一化估计的激活统计数据，一致地提高了 25 种不同流行计算机视觉模型的鲁棒性。使用校正后的统计数据，ResNet-50 在 ImageNet-C 上达到 62.2% mCE，而 76。7% 无适应。借助更强大的 DeepAugment+AugMix 模型，我们将 ResNet50 模型实现的最新技术从 53.6% mCE 提高到 45.4% mCE。即使适应单个样本也可以提高 ResNet-50 和 AugMix 模型的鲁棒性，32 个样本足以改进 ResNet-50 架构的当前技术水平。我们认为，无论何时报告腐败基准和其他分布外泛化设置中的分数，都应包括经过调整的统计结果。32 个样本足以改进 ResNet-50 架构的当前技术水平。我们认为，无论何时报告腐败基准和其他分布外泛化设置中的分数，都应包括经过调整的统计结果。32 个样本足以改进 ResNet-50 架构的当前技术水平。我们认为，无论何时报告腐败基准和其他分布外泛化设置中的分数，都应包括经过调整的统计结果。