Matches in SemOpenAlex for { <https://semopenalex.org/work/W4386065611> ?p ?o ?g. }
Showing items 1 to 95 of
95
with 100 items per page.
- W4386065611 abstract "The task of Visual Question Answering (VQA) is known to be plagued by the issue of VQA models exploiting biases within the dataset to make its final prediction. Various previous ensemble based debiasing methods have been proposed where an additional model is purposefully trained to be biased in order to train a robust target model. However, these methods compute the bias for a model simply from the label statistics of the training data or from single modal branches. In this work, in order to better learn the bias a target VQA model suffers from, we propose a generative method to train the bias model directly from the target model, called GenB. In particular, GenB employs a generative network to learn the bias in the target model through a combination of the adversarial objective and knowledge distillation. We then debias our target model with GenB as a bias model, and show through extensive experiments the effects of our method on various VQA bias datasets including VQA-CP2, VQA-CP1, GQA-OOD, and VQA-CE, and show state-of-the-art results with the LXMERT architecture on VQA-CP2." @default.
- W4386065611 created "2023-08-23" @default.
- W4386065611 creator A5012455275 @default.
- W4386065611 creator A5014059162 @default.
- W4386065611 creator A5047089963 @default.
- W4386065611 creator A5056053341 @default.
- W4386065611 date "2023-06-01" @default.
- W4386065611 modified "2023-10-06" @default.
- W4386065611 title "Generative Bias for Robust Visual Question Answering" @default.
- W4386065611 cites W1933349210 @default.
- W4386065611 cites W2250539671 @default.
- W4386065611 cites W2560730294 @default.
- W4386065611 cites W2561715562 @default.
- W4386065611 cites W2745461083 @default.
- W4386065611 cites W2963073614 @default.
- W4386065611 cites W2963260436 @default.
- W4386065611 cites W2963622213 @default.
- W4386065611 cites W2963880187 @default.
- W4386065611 cites W2963890019 @default.
- W4386065611 cites W2963954913 @default.
- W4386065611 cites W2970019270 @default.
- W4386065611 cites W2970231061 @default.
- W4386065611 cites W2976658881 @default.
- W4386065611 cites W2983256121 @default.
- W4386065611 cites W3035145964 @default.
- W4386065611 cites W3035517717 @default.
- W4386065611 cites W3103211586 @default.
- W4386065611 cites W3104788521 @default.
- W4386065611 cites W3174366544 @default.
- W4386065611 cites W3177934633 @default.
- W4386065611 cites W3186953717 @default.
- W4386065611 cites W3201957104 @default.
- W4386065611 cites W3202778561 @default.
- W4386065611 cites W3205625102 @default.
- W4386065611 cites W4312300358 @default.
- W4386065611 cites W4312373370 @default.
- W4386065611 cites W4312457137 @default.
- W4386065611 cites W4372267434 @default.
- W4386065611 cites W3199267905 @default.
- W4386065611 doi "https://doi.org/10.1109/cvpr52729.2023.01124" @default.
- W4386065611 hasPublicationYear "2023" @default.
- W4386065611 type Work @default.
- W4386065611 citedByCount "2" @default.
- W4386065611 countsByYear W43860656112023 @default.
- W4386065611 crossrefType "proceedings-article" @default.
- W4386065611 hasAuthorship W4386065611A5012455275 @default.
- W4386065611 hasAuthorship W4386065611A5014059162 @default.
- W4386065611 hasAuthorship W4386065611A5047089963 @default.
- W4386065611 hasAuthorship W4386065611A5056053341 @default.
- W4386065611 hasConcept C119857082 @default.
- W4386065611 hasConcept C124101348 @default.
- W4386065611 hasConcept C154945302 @default.
- W4386065611 hasConcept C15744967 @default.
- W4386065611 hasConcept C162324750 @default.
- W4386065611 hasConcept C167966045 @default.
- W4386065611 hasConcept C187736073 @default.
- W4386065611 hasConcept C188147891 @default.
- W4386065611 hasConcept C2779458634 @default.
- W4386065611 hasConcept C2780451532 @default.
- W4386065611 hasConcept C39890363 @default.
- W4386065611 hasConcept C41008148 @default.
- W4386065611 hasConcept C44291984 @default.
- W4386065611 hasConcept C97256817 @default.
- W4386065611 hasConceptScore W4386065611C119857082 @default.
- W4386065611 hasConceptScore W4386065611C124101348 @default.
- W4386065611 hasConceptScore W4386065611C154945302 @default.
- W4386065611 hasConceptScore W4386065611C15744967 @default.
- W4386065611 hasConceptScore W4386065611C162324750 @default.
- W4386065611 hasConceptScore W4386065611C167966045 @default.
- W4386065611 hasConceptScore W4386065611C187736073 @default.
- W4386065611 hasConceptScore W4386065611C188147891 @default.
- W4386065611 hasConceptScore W4386065611C2779458634 @default.
- W4386065611 hasConceptScore W4386065611C2780451532 @default.
- W4386065611 hasConceptScore W4386065611C39890363 @default.
- W4386065611 hasConceptScore W4386065611C41008148 @default.
- W4386065611 hasConceptScore W4386065611C44291984 @default.
- W4386065611 hasConceptScore W4386065611C97256817 @default.
- W4386065611 hasFunder F4320321142 @default.
- W4386065611 hasFunder F4320324891 @default.
- W4386065611 hasLocation W43860656111 @default.
- W4386065611 hasOpenAccess W4386065611 @default.
- W4386065611 hasPrimaryLocation W43860656111 @default.
- W4386065611 hasRelatedWork W1534961803 @default.
- W4386065611 hasRelatedWork W2952072295 @default.
- W4386065611 hasRelatedWork W3206427007 @default.
- W4386065611 hasRelatedWork W4221160981 @default.
- W4386065611 hasRelatedWork W4289760695 @default.
- W4386065611 hasRelatedWork W4297802107 @default.
- W4386065611 hasRelatedWork W4320516002 @default.
- W4386065611 hasRelatedWork W4385572368 @default.
- W4386065611 hasRelatedWork W4385573187 @default.
- W4386065611 hasRelatedWork W2310403681 @default.
- W4386065611 isParatext "false" @default.
- W4386065611 isRetracted "false" @default.
- W4386065611 workType "article" @default.