Matches in SemOpenAlex for { <https://semopenalex.org/work/W4287183747> ?p ?o ?g. }
Showing items 1 to 67 of
67
with 100 items per page.
- W4287183747 abstract "Leveraging the characteristics of convolutional layers, neural networks are extremely effective for pattern recognition tasks. However in some cases, their decisions are based on unintended information leading to high performance on standard benchmarks but also to a lack of generalization to challenging testing conditions and unintuitive failures. Recent work has termed this shortcut learning and addressed its presence in multiple domains. In text recognition, we reveal another such shortcut, whereby recognizers overly depend on local image statistics. Motivated by this, we suggest an approach to regulate the reliance on local statistics that improves text recognition performance. Our method, termed TextAdaIN, creates local distortions in the feature map which prevent the network from overfitting to local statistics. It does so by viewing each feature map as a sequence of elements and deliberately mismatching fine-grained feature statistics between elements in a mini-batch. Despite TextAdaIN's simplicity, extensive experiments show its effectiveness compared to other, more complicated methods. TextAdaIN achieves state-of-the-art results on standard handwritten text recognition benchmarks. It generalizes to multiple architectures and to the domain of scene text recognition. Furthermore, we demonstrate that integrating TextAdaIN improves robustness towards more challenging testing conditions. The official Pytorch implementation can be found at https://github.com/amazon-research/textadain-robust-recognition." @default.
- W4287183747 created "2022-07-25" @default.
- W4287183747 creator A5048711507 @default.
- W4287183747 creator A5053244815 @default.
- W4287183747 creator A5075513070 @default.
- W4287183747 date "2021-05-09" @default.
- W4287183747 modified "2023-09-23" @default.
- W4287183747 title "TextAdaIN: Paying Attention to Shortcut Learning in Text Recognizers" @default.
- W4287183747 doi "https://doi.org/10.48550/arxiv.2105.03906" @default.
- W4287183747 hasPublicationYear "2021" @default.
- W4287183747 type Work @default.
- W4287183747 citedByCount "0" @default.
- W4287183747 crossrefType "posted-content" @default.
- W4287183747 hasAuthorship W4287183747A5048711507 @default.
- W4287183747 hasAuthorship W4287183747A5053244815 @default.
- W4287183747 hasAuthorship W4287183747A5075513070 @default.
- W4287183747 hasBestOaLocation W42871837471 @default.
- W4287183747 hasConcept C104317684 @default.
- W4287183747 hasConcept C119857082 @default.
- W4287183747 hasConcept C134306372 @default.
- W4287183747 hasConcept C138885662 @default.
- W4287183747 hasConcept C153180895 @default.
- W4287183747 hasConcept C154945302 @default.
- W4287183747 hasConcept C177148314 @default.
- W4287183747 hasConcept C185592680 @default.
- W4287183747 hasConcept C22019652 @default.
- W4287183747 hasConcept C2776401178 @default.
- W4287183747 hasConcept C33923547 @default.
- W4287183747 hasConcept C41008148 @default.
- W4287183747 hasConcept C41895202 @default.
- W4287183747 hasConcept C50644808 @default.
- W4287183747 hasConcept C55493867 @default.
- W4287183747 hasConcept C63479239 @default.
- W4287183747 hasConcept C81363708 @default.
- W4287183747 hasConceptScore W4287183747C104317684 @default.
- W4287183747 hasConceptScore W4287183747C119857082 @default.
- W4287183747 hasConceptScore W4287183747C134306372 @default.
- W4287183747 hasConceptScore W4287183747C138885662 @default.
- W4287183747 hasConceptScore W4287183747C153180895 @default.
- W4287183747 hasConceptScore W4287183747C154945302 @default.
- W4287183747 hasConceptScore W4287183747C177148314 @default.
- W4287183747 hasConceptScore W4287183747C185592680 @default.
- W4287183747 hasConceptScore W4287183747C22019652 @default.
- W4287183747 hasConceptScore W4287183747C2776401178 @default.
- W4287183747 hasConceptScore W4287183747C33923547 @default.
- W4287183747 hasConceptScore W4287183747C41008148 @default.
- W4287183747 hasConceptScore W4287183747C41895202 @default.
- W4287183747 hasConceptScore W4287183747C50644808 @default.
- W4287183747 hasConceptScore W4287183747C55493867 @default.
- W4287183747 hasConceptScore W4287183747C63479239 @default.
- W4287183747 hasConceptScore W4287183747C81363708 @default.
- W4287183747 hasLocation W42871837471 @default.
- W4287183747 hasOpenAccess W4287183747 @default.
- W4287183747 hasPrimaryLocation W42871837471 @default.
- W4287183747 hasRelatedWork W11538522 @default.
- W4287183747 hasRelatedWork W12783365 @default.
- W4287183747 hasRelatedWork W13188192 @default.
- W4287183747 hasRelatedWork W14789944 @default.
- W4287183747 hasRelatedWork W3092890 @default.
- W4287183747 hasRelatedWork W6229082 @default.
- W4287183747 hasRelatedWork W6882942 @default.
- W4287183747 hasRelatedWork W9190101 @default.
- W4287183747 hasRelatedWork W9402503 @default.
- W4287183747 hasRelatedWork W9940185 @default.
- W4287183747 isParatext "false" @default.
- W4287183747 isRetracted "false" @default.
- W4287183747 workType "article" @default.