{"id":118504,"date":"2021-10-31T23:47:16","date_gmt":"2021-10-31T14:47:16","guid":{"rendered":"http:\/\/175.125.95.178\/ai-in-signal\/18504\/"},"modified":"2026-04-27T16:12:04","modified_gmt":"2026-04-27T07:12:04","slug":"18504","status":"publish","type":"ai-in-signal","link":"http:\/\/ee.presscat.kr\/en\/ai-in-signal\/18504\/","title":{"rendered":"Revisiting Batch Normalization for Improving Corruption Robustness (Prof. In-So Kweon)"},"content":{"rendered":"<p style=\"text-align:justify;margin-bottom:11px\"><span style=\"font-size:10pt\"><span style=\"line-height:107%\"><span>Conference\/Journal, Year: WACV 2021<\/span><\/span><\/span><\/p>\n<p style=\"text-align:justify;margin-bottom:11px\"><span style=\"font-size:10pt\"><span style=\"line-height:107%\"><span>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%.<\/span><\/span><\/span><\/p>\n<p style=\"text-align:justify;margin-bottom:11px\"><span style=\"font-size:10pt\"><span style=\"line-height:107%\"><span><\/p>\n<div class=\"\"><img decoding=\"async\" class=\"\" src=\"\/wp-content\/uploads\/drupal\/\uad8c\uc778\uc18c\uad50\uc218\ub2d823.png\" alt=\"\" title=\"\"><\/div>\n<p><\/span><\/span><\/span><\/p>\n","protected":false},"excerpt":{"rendered":"<p>620<\/p>\n","protected":false},"featured_media":0,"template":"","class_list":["post-118504","ai-in-signal","type-ai-in-signal","status-publish","hentry"],"acf":[],"_links":{"self":[{"href":"http:\/\/ee.presscat.kr\/en\/wp-json\/wp\/v2\/ai-in-signal\/118504","targetHints":{"allow":["GET"]}}],"collection":[{"href":"http:\/\/ee.presscat.kr\/en\/wp-json\/wp\/v2\/ai-in-signal"}],"about":[{"href":"http:\/\/ee.presscat.kr\/en\/wp-json\/wp\/v2\/types\/ai-in-signal"}],"wp:attachment":[{"href":"http:\/\/ee.presscat.kr\/en\/wp-json\/wp\/v2\/media?parent=118504"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}