Matches in SemOpenAlex for { <https://semopenalex.org/work/W4385488514> ?p ?o ?g. }
Showing items 1 to 78 of
78
with 100 items per page.
- W4385488514 abstract "Deep neural networks achieve high prediction accuracy when the train and test distributions coincide. In practice though, various types of corruptions occur which deviate from this setup and cause severe performance degradations. Few methods have been proposed to address generalization in the presence of unforeseen domain shifts. In particular, digital noise corruptions arise commonly in practice during the image acquisition stage and present a significant challenge for current methods. In this paper, we propose a diverse Gaussian noise consistency regularization method for improving robustness of image classifiers under a variety of corruptions while still maintaining high clean accuracy. We derive bounds to motivate and understand the behavior of our Gaussian noise consistency regularization using a local loss landscape analysis. Our approach improves robustness against unforeseen noise corruptions by 4.2-18.4 % over adversarial training and other strong diverse data augmentation baselines across several benchmarks. Furthermore, it improves robustness and uncertainty calibration by 3.7% and 5.5 %, respectively, against all common corruptions (weather, digital, blur, noise) when combined with state-of-the-art diverse data augmentations. Code is available at https://github.com/TheoT1/DiGN." @default.
- W4385488514 created "2023-08-03" @default.
- W4385488514 creator A5031789494 @default.
- W4385488514 creator A5088931370 @default.
- W4385488514 date "2023-06-18" @default.
- W4385488514 modified "2023-09-26" @default.
- W4385488514 title "Diverse Gaussian Noise Consistency Regularization for Robustness and Uncertainty Calibration" @default.
- W4385488514 cites W1677182931 @default.
- W4385488514 cites W2031342017 @default.
- W4385488514 cites W2732452266 @default.
- W4385488514 cites W2768346313 @default.
- W4385488514 cites W2913496322 @default.
- W4385488514 cites W2949736877 @default.
- W4385488514 cites W2962710014 @default.
- W4385488514 cites W2963026800 @default.
- W4385488514 cites W2963031676 @default.
- W4385488514 cites W2970121940 @default.
- W4385488514 cites W2985282977 @default.
- W4385488514 cites W2992308087 @default.
- W4385488514 cites W3037492894 @default.
- W4385488514 cites W3214886976 @default.
- W4385488514 doi "https://doi.org/10.1109/ijcnn54540.2023.10191763" @default.
- W4385488514 hasPublicationYear "2023" @default.
- W4385488514 type Work @default.
- W4385488514 citedByCount "0" @default.
- W4385488514 crossrefType "proceedings-article" @default.
- W4385488514 hasAuthorship W4385488514A5031789494 @default.
- W4385488514 hasAuthorship W4385488514A5088931370 @default.
- W4385488514 hasBestOaLocation W43854885142 @default.
- W4385488514 hasConcept C104317684 @default.
- W4385488514 hasConcept C11413529 @default.
- W4385488514 hasConcept C119857082 @default.
- W4385488514 hasConcept C121332964 @default.
- W4385488514 hasConcept C124101348 @default.
- W4385488514 hasConcept C154945302 @default.
- W4385488514 hasConcept C163294075 @default.
- W4385488514 hasConcept C163716315 @default.
- W4385488514 hasConcept C185592680 @default.
- W4385488514 hasConcept C2776135515 @default.
- W4385488514 hasConcept C29265498 @default.
- W4385488514 hasConcept C41008148 @default.
- W4385488514 hasConcept C4199805 @default.
- W4385488514 hasConcept C55493867 @default.
- W4385488514 hasConcept C62520636 @default.
- W4385488514 hasConcept C63479239 @default.
- W4385488514 hasConceptScore W4385488514C104317684 @default.
- W4385488514 hasConceptScore W4385488514C11413529 @default.
- W4385488514 hasConceptScore W4385488514C119857082 @default.
- W4385488514 hasConceptScore W4385488514C121332964 @default.
- W4385488514 hasConceptScore W4385488514C124101348 @default.
- W4385488514 hasConceptScore W4385488514C154945302 @default.
- W4385488514 hasConceptScore W4385488514C163294075 @default.
- W4385488514 hasConceptScore W4385488514C163716315 @default.
- W4385488514 hasConceptScore W4385488514C185592680 @default.
- W4385488514 hasConceptScore W4385488514C2776135515 @default.
- W4385488514 hasConceptScore W4385488514C29265498 @default.
- W4385488514 hasConceptScore W4385488514C41008148 @default.
- W4385488514 hasConceptScore W4385488514C4199805 @default.
- W4385488514 hasConceptScore W4385488514C55493867 @default.
- W4385488514 hasConceptScore W4385488514C62520636 @default.
- W4385488514 hasConceptScore W4385488514C63479239 @default.
- W4385488514 hasLocation W43854885141 @default.
- W4385488514 hasLocation W43854885142 @default.
- W4385488514 hasOpenAccess W4385488514 @default.
- W4385488514 hasPrimaryLocation W43854885141 @default.
- W4385488514 hasRelatedWork W1513142184 @default.
- W4385488514 hasRelatedWork W1648249948 @default.
- W4385488514 hasRelatedWork W1984740845 @default.
- W4385488514 hasRelatedWork W2087200833 @default.
- W4385488514 hasRelatedWork W2130950408 @default.
- W4385488514 hasRelatedWork W2169600804 @default.
- W4385488514 hasRelatedWork W2805046202 @default.
- W4385488514 hasRelatedWork W2948805184 @default.
- W4385488514 hasRelatedWork W3115334024 @default.
- W4385488514 hasRelatedWork W4288335398 @default.
- W4385488514 isParatext "false" @default.
- W4385488514 isRetracted "false" @default.
- W4385488514 workType "article" @default.