Matches in SemOpenAlex for { <https://semopenalex.org/work/W4285345750> ?p ?o ?g. }
Showing items 1 to 76 of
76
with 100 items per page.
- W4285345750 abstract "BERT-type structure has led to the revolution of vision-language pre-training and the achievement of state-of-the-art results on numerous vision-language downstream tasks. Existing solutions dominantly capitalize on the multi-modal inputs with mask tokens to trigger mask-based proxy pre-training tasks (e.g., masked language modeling and masked object/frame prediction). In this work, we argue that such masked inputs would inevitably introduce noise for cross-modal matching proxy task, and thus leave the inherent vision-language association under-explored. As an alternative, we derive a particular form of cross-modal proxy objective for video-language pre-training, i.e., Contrastive Cross-modal matching and denoising (CoCo). By viewing the masked frame/word sequences as the noisy augmentation of primary unmasked ones, CoCo strengthens video-language association by simultaneously pursuing inter-modal matching and intra-modal denoising between masked and unmasked inputs in a contrastive manner. Our CoCo proxy objective can be further integrated into any BERT-type encoder-decoder structure for video-language pre-training, named as Contrastive Cross-modal BERT (CoCo-BERT). We pre-train CoCo-BERT on TV dataset and a newly collected large-scale GIF video dataset (ACTION). Through extensive experiments over a wide range of downstream tasks (e.g., cross-modal retrieval, video question answering, and video captioning), we demonstrate the superiority of CoCo-BERT as a pre-trained structure." @default.
- W4285345750 created "2022-07-14" @default.
- W4285345750 creator A5007558888 @default.
- W4285345750 creator A5017597537 @default.
- W4285345750 creator A5041154840 @default.
- W4285345750 creator A5061525421 @default.
- W4285345750 creator A5085403640 @default.
- W4285345750 creator A5088760097 @default.
- W4285345750 date "2021-10-17" @default.
- W4285345750 modified "2023-09-25" @default.
- W4285345750 title "CoCo-BERT" @default.
- W4285345750 cites W1586939924 @default.
- W4285345750 cites W1995820507 @default.
- W4285345750 cites W2108598243 @default.
- W4285345750 cites W2138621090 @default.
- W4285345750 cites W2766375149 @default.
- W4285345750 cites W2808399042 @default.
- W4285345750 cites W2895845501 @default.
- W4285345750 cites W2962907269 @default.
- W4285345750 cites W2963971014 @default.
- W4285345750 cites W2983141445 @default.
- W4285345750 cites W2990503944 @default.
- W4285345750 cites W2997591391 @default.
- W4285345750 cites W3035365026 @default.
- W4285345750 cites W3090449556 @default.
- W4285345750 doi "https://doi.org/10.1145/3474085.3475703" @default.
- W4285345750 hasPublicationYear "2021" @default.
- W4285345750 type Work @default.
- W4285345750 citedByCount "16" @default.
- W4285345750 countsByYear W42853457502022 @default.
- W4285345750 countsByYear W42853457502023 @default.
- W4285345750 crossrefType "proceedings-article" @default.
- W4285345750 hasAuthorship W4285345750A5007558888 @default.
- W4285345750 hasAuthorship W4285345750A5017597537 @default.
- W4285345750 hasAuthorship W4285345750A5041154840 @default.
- W4285345750 hasAuthorship W4285345750A5061525421 @default.
- W4285345750 hasAuthorship W4285345750A5085403640 @default.
- W4285345750 hasAuthorship W4285345750A5088760097 @default.
- W4285345750 hasConcept C111919701 @default.
- W4285345750 hasConcept C115961682 @default.
- W4285345750 hasConcept C118505674 @default.
- W4285345750 hasConcept C154945302 @default.
- W4285345750 hasConcept C157657479 @default.
- W4285345750 hasConcept C185592680 @default.
- W4285345750 hasConcept C188027245 @default.
- W4285345750 hasConcept C204321447 @default.
- W4285345750 hasConcept C28490314 @default.
- W4285345750 hasConcept C41008148 @default.
- W4285345750 hasConcept C71139939 @default.
- W4285345750 hasConceptScore W4285345750C111919701 @default.
- W4285345750 hasConceptScore W4285345750C115961682 @default.
- W4285345750 hasConceptScore W4285345750C118505674 @default.
- W4285345750 hasConceptScore W4285345750C154945302 @default.
- W4285345750 hasConceptScore W4285345750C157657479 @default.
- W4285345750 hasConceptScore W4285345750C185592680 @default.
- W4285345750 hasConceptScore W4285345750C188027245 @default.
- W4285345750 hasConceptScore W4285345750C204321447 @default.
- W4285345750 hasConceptScore W4285345750C28490314 @default.
- W4285345750 hasConceptScore W4285345750C41008148 @default.
- W4285345750 hasConceptScore W4285345750C71139939 @default.
- W4285345750 hasLocation W42853457501 @default.
- W4285345750 hasOpenAccess W4285345750 @default.
- W4285345750 hasPrimaryLocation W42853457501 @default.
- W4285345750 hasRelatedWork W2503073734 @default.
- W4285345750 hasRelatedWork W2547835662 @default.
- W4285345750 hasRelatedWork W2596543464 @default.
- W4285345750 hasRelatedWork W2891852518 @default.
- W4285345750 hasRelatedWork W2905654560 @default.
- W4285345750 hasRelatedWork W2923366293 @default.
- W4285345750 hasRelatedWork W3008515501 @default.
- W4285345750 hasRelatedWork W3183824823 @default.
- W4285345750 hasRelatedWork W4320016117 @default.
- W4285345750 hasRelatedWork W2519434724 @default.
- W4285345750 isParatext "false" @default.
- W4285345750 isRetracted "false" @default.
- W4285345750 workType "article" @default.