Dynamic Batch Norm Statistics Update for Natural Robustness
This addresses the issue of poor performance of off-the-shelf trained models on corrupted images, offering a method that works without retraining, though it is incremental as it builds on prior BN statistics update techniques.
The paper tackles the problem of improving DNN robustness to unknown image corruptions at inference time by proposing a unified framework that detects corruption types in the Fourier domain and updates BatchNorm statistics, achieving about 8% and 4% accuracy improvements on CIFAR10-C and ImageNet-C, respectively.
DNNs trained on natural clean samples have been shown to perform poorly on corrupted samples, such as noisy or blurry images. Various data augmentation methods have been recently proposed to improve DNN's robustness against common corruptions. Despite their success, they require computationally expensive training and cannot be applied to off-the-shelf trained models. Recently, it has been shown that updating BatchNorm (BN) statistics of an off-the-shelf model on a single corruption improves its accuracy on that corruption significantly. However, adopting the idea at inference time when the type of corruption is unknown and changing decreases the effectiveness of this method. In this paper, we harness the Fourier domain to detect the corruption type, a challenging task in the image domain. We propose a unified framework consisting of a corruption-detection model and BN statistics update that improves the corruption accuracy of any off-the-shelf trained model. We benchmark our framework on different models and datasets. Our results demonstrate about 8% and 4% accuracy improvement on CIFAR10-C and ImageNet-C, respectively. Furthermore, our framework can further improve the accuracy of state-of-the-art robust models, such as AugMix and DeepAug.