Matches in SemOpenAlex for { <https://semopenalex.org/work/W3100137005> ?p ?o ?g. }
Showing items 1 to 91 of
91
with 100 items per page.
- W3100137005 abstract "The success of learning with noisy labels (LNL) methods relies heavily on the success of a warm-up stage where standard supervised training is performed using the full (noisy) training set. In this paper, we identify a warm-up obstacle: the inability of standard warm-up stages to train high quality feature extractors and avert memorization of noisy labels. We propose Contrast to Divide (C2D), a simple framework that solves this problem by pre-training the feature extractor in a self-supervised fashion. Using self-supervised pre-training boosts the performance of existing LNL approaches by drastically reducing the warm-up stage's susceptibility to noise level, shortening its duration, and improving extracted feature quality. C2D works out of the box with existing methods and demonstrates markedly improved performance, especially in the high noise regime, where we get a boost of more than 27% for CIFAR-100 with 90% noise over the previous state of the art. In real-life noise settings, C2D trained on mini-WebVision outperforms previous works both in WebVision and ImageNet validation sets by 3% top-1 accuracy. We perform an in-depth analysis of the framework, including investigating the performance of different pre-training approaches and estimating the effective upper bound of the LNL performance with semi-supervised learning. Code for reproducing our experiments is available at https://github.com/ContrastToDivide/C2D." @default.
- W3100137005 created "2020-11-23" @default.
- W3100137005 creator A5006347979 @default.
- W3100137005 creator A5019913171 @default.
- W3100137005 creator A5025410418 @default.
- W3100137005 creator A5029515956 @default.
- W3100137005 creator A5089135250 @default.
- W3100137005 date "2022-01-01" @default.
- W3100137005 modified "2023-10-03" @default.
- W3100137005 title "Contrast to Divide: Self-Supervised Pre-Training for Learning with Noisy Labels" @default.
- W3100137005 cites W2025768430 @default.
- W3100137005 cites W2117539524 @default.
- W3100137005 cites W2138621090 @default.
- W3100137005 cites W2321533354 @default.
- W3100137005 cites W2558661413 @default.
- W3100137005 cites W2577784528 @default.
- W3100137005 cites W2799056151 @default.
- W3100137005 cites W2913939497 @default.
- W3100137005 cites W2948606739 @default.
- W3100137005 cites W2963420272 @default.
- W3100137005 cites W2963703197 @default.
- W3100137005 cites W2964155802 @default.
- W3100137005 cites W2964274690 @default.
- W3100137005 cites W2967052791 @default.
- W3100137005 cites W2978625989 @default.
- W3100137005 cites W2981952041 @default.
- W3100137005 cites W2990019157 @default.
- W3100137005 cites W3034781633 @default.
- W3100137005 cites W343636949 @default.
- W3100137005 cites W4288083516 @default.
- W3100137005 doi "https://doi.org/10.1109/wacv51458.2022.00046" @default.
- W3100137005 hasPublicationYear "2022" @default.
- W3100137005 type Work @default.
- W3100137005 sameAs 3100137005 @default.
- W3100137005 citedByCount "21" @default.
- W3100137005 countsByYear W31001370052021 @default.
- W3100137005 countsByYear W31001370052022 @default.
- W3100137005 countsByYear W31001370052023 @default.
- W3100137005 crossrefType "proceedings-article" @default.
- W3100137005 hasAuthorship W3100137005A5006347979 @default.
- W3100137005 hasAuthorship W3100137005A5019913171 @default.
- W3100137005 hasAuthorship W3100137005A5025410418 @default.
- W3100137005 hasAuthorship W3100137005A5029515956 @default.
- W3100137005 hasAuthorship W3100137005A5089135250 @default.
- W3100137005 hasBestOaLocation W31001370052 @default.
- W3100137005 hasConcept C115961682 @default.
- W3100137005 hasConcept C119857082 @default.
- W3100137005 hasConcept C136389625 @default.
- W3100137005 hasConcept C138885662 @default.
- W3100137005 hasConcept C153180895 @default.
- W3100137005 hasConcept C154945302 @default.
- W3100137005 hasConcept C177264268 @default.
- W3100137005 hasConcept C199360897 @default.
- W3100137005 hasConcept C2776401178 @default.
- W3100137005 hasConcept C2776502983 @default.
- W3100137005 hasConcept C41008148 @default.
- W3100137005 hasConcept C41895202 @default.
- W3100137005 hasConcept C50644808 @default.
- W3100137005 hasConcept C99498987 @default.
- W3100137005 hasConceptScore W3100137005C115961682 @default.
- W3100137005 hasConceptScore W3100137005C119857082 @default.
- W3100137005 hasConceptScore W3100137005C136389625 @default.
- W3100137005 hasConceptScore W3100137005C138885662 @default.
- W3100137005 hasConceptScore W3100137005C153180895 @default.
- W3100137005 hasConceptScore W3100137005C154945302 @default.
- W3100137005 hasConceptScore W3100137005C177264268 @default.
- W3100137005 hasConceptScore W3100137005C199360897 @default.
- W3100137005 hasConceptScore W3100137005C2776401178 @default.
- W3100137005 hasConceptScore W3100137005C2776502983 @default.
- W3100137005 hasConceptScore W3100137005C41008148 @default.
- W3100137005 hasConceptScore W3100137005C41895202 @default.
- W3100137005 hasConceptScore W3100137005C50644808 @default.
- W3100137005 hasConceptScore W3100137005C99498987 @default.
- W3100137005 hasLocation W31001370051 @default.
- W3100137005 hasLocation W31001370052 @default.
- W3100137005 hasOpenAccess W3100137005 @default.
- W3100137005 hasPrimaryLocation W31001370051 @default.
- W3100137005 hasRelatedWork W2015538044 @default.
- W3100137005 hasRelatedWork W2016461833 @default.
- W3100137005 hasRelatedWork W2052253960 @default.
- W3100137005 hasRelatedWork W2382607599 @default.
- W3100137005 hasRelatedWork W2760085659 @default.
- W3100137005 hasRelatedWork W2970216048 @default.
- W3100137005 hasRelatedWork W3126323604 @default.
- W3100137005 hasRelatedWork W3197541072 @default.
- W3100137005 hasRelatedWork W376702462 @default.
- W3100137005 hasRelatedWork W2480412556 @default.
- W3100137005 isParatext "false" @default.
- W3100137005 isRetracted "false" @default.
- W3100137005 magId "3100137005" @default.
- W3100137005 workType "article" @default.