Revisiting Batch Normalization for Improving Corruption Robustness
The performance of DNNs trained on clean images has been shown to decrease when the test images have common corruptions. In this work, we interpret corruption robustness as a domain shift and propose to rectify batch normalization (BN) statistics for improving model robustness. This is motivated by perceiving the shift from the clean domain to the corruption domain as a style shift that is represented by the BN statistics. We find that simply estimating and adapting the BN statistics on a few (32 for instance) representation samples, without retraining the model, improves the corruption robustness by a large margin on several benchmark datasets with a wide range of model architectures. For example, on ImageNet-C, statistics adaptation improves the top1 accuracy of ResNet50 from 39.2% to 48.7%. Moreover, we find that this technique can further improve state-of-the-art robust models from 58.1% to 63.3%.
Code (0)
등록된 구현이 없습니다.
Methods 이 논문이 사용한 방법론
Similar Papers 제목 키워드 기반
Distribution Mismatch Correction for Improved Robustness in Deep Neural Networks
Deep neural networks rely heavily on normalization methods to improve their performance and learning behavior. Although normalization methods spurred the development of increasingly deep and efficient architectures, they…
Benchmarking the Robustness of Instance Segmentation Models
This paper presents a comprehensive evaluation of instance segmentation models with respect to real-world image corruptions as well as out-of-domain image collections, e.g. images captured by a different set-up than the …
BenchmarkingDomain AdaptationInstance SegmentationSegmentation+1Batch Normalization Provably Avoids Rank Collapse for Randomly Initialised Deep Networks
Randomly initialized neural networks are known to become harder to train with increasing depth, unless architectural enhancements like residual connections and batch normalization are used. We here investigate this pheno…
Batch normalization provably avoids ranks collapse for randomly initialised deep networks
Randomly initialized neural networks are known to become harder to train with increasing depth, unless architectural enhancements like residual connections and batch normalization are used. We here investigate this pheno…
Improving robustness against common corruptions by covariate shift adaptation
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…
Domain GeneralizationImage ClassificationOut-of-Distribution GeneralizationUnsupervised Domain Adaptation