Matches in SemOpenAlex for { <https://semopenalex.org/work/W3109771530> ?p ?o ?g. }
- W3109771530 endingPage "5033" @default.
- W3109771530 startingPage "5025" @default.
- W3109771530 abstract "A major challenge in scene graph classification is that the appearance of objects and relations can be significantly different from one image to another. Previous works have addressed this by relational reasoning over all objects in an image or incorporating prior knowledge into classification. Unlike previous works, we do not consider separate models for perception and prior knowledge. Instead, we take a multi-task learning approach by introducing schema representations and implementing the classification as an attention layer between image-based representations and the schemata. This allows for the prior knowledge to emerge and propagate within the perception model. By enforcing the model also to represent the prior, we achieve a strong inductive bias. We show that our model can accurately generate commonsense knowledge and that the iterative injection of this knowledge to scene representations, as a top-down mechanism, leads to significantly higher classification performance. Additionally, our model can be fine-tuned on external knowledge given as triples. When combined with self-supervised learning and with 1% of annotated images only, this gives more than 3% improvement in object classification, 26% in scene graph classification, and 36% in predicate prediction accuracy." @default.
- W3109771530 created "2020-12-07" @default.
- W3109771530 creator A5039442998 @default.
- W3109771530 creator A5074808403 @default.
- W3109771530 creator A5081679022 @default.
- W3109771530 date "2021-05-18" @default.
- W3109771530 modified "2023-10-16" @default.
- W3109771530 title "Classification by Attention: Scene Graph Classification with Prior Knowledge" @default.
- W3109771530 cites W1522301498 @default.
- W3109771530 cites W1529533208 @default.
- W3109771530 cites W1533861849 @default.
- W3109771530 cites W1686810756 @default.
- W3109771530 cites W205829674 @default.
- W3109771530 cites W2064675550 @default.
- W3109771530 cites W2108598243 @default.
- W3109771530 cites W2118373646 @default.
- W3109771530 cites W2127795553 @default.
- W3109771530 cites W2161388792 @default.
- W3109771530 cites W2187089797 @default.
- W3109771530 cites W2194775991 @default.
- W3109771530 cites W2277195237 @default.
- W3109771530 cites W2295647632 @default.
- W3109771530 cites W2475771773 @default.
- W3109771530 cites W2479423890 @default.
- W3109771530 cites W2519887557 @default.
- W3109771530 cites W2579549467 @default.
- W3109771530 cites W2591644541 @default.
- W3109771530 cites W2604603079 @default.
- W3109771530 cites W2613718673 @default.
- W3109771530 cites W2624780181 @default.
- W3109771530 cites W2759710969 @default.
- W3109771530 cites W2763058042 @default.
- W3109771530 cites W2781474777 @default.
- W3109771530 cites W2805516822 @default.
- W3109771530 cites W2827853180 @default.
- W3109771530 cites W2886970679 @default.
- W3109771530 cites W2913340405 @default.
- W3109771530 cites W2914694623 @default.
- W3109771530 cites W2941109201 @default.
- W3109771530 cites W2943248951 @default.
- W3109771530 cites W2948818258 @default.
- W3109771530 cites W2954922414 @default.
- W3109771530 cites W2962737704 @default.
- W3109771530 cites W2963403868 @default.
- W3109771530 cites W2963513598 @default.
- W3109771530 cites W2963536419 @default.
- W3109771530 cites W2963907629 @default.
- W3109771530 cites W2963938081 @default.
- W3109771530 cites W2970523217 @default.
- W3109771530 cites W2982112268 @default.
- W3109771530 cites W2985775525 @default.
- W3109771530 cites W2987123286 @default.
- W3109771530 cites W3003588813 @default.
- W3109771530 cites W3035060554 @default.
- W3109771530 cites W3037655549 @default.
- W3109771530 cites W3091156754 @default.
- W3109771530 cites W64813323 @default.
- W3109771530 cites W648786980 @default.
- W3109771530 doi "https://doi.org/10.1609/aaai.v35i6.16636" @default.
- W3109771530 hasPublicationYear "2021" @default.
- W3109771530 type Work @default.
- W3109771530 sameAs 3109771530 @default.
- W3109771530 citedByCount "13" @default.
- W3109771530 countsByYear W31097715302021 @default.
- W3109771530 countsByYear W31097715302022 @default.
- W3109771530 countsByYear W31097715302023 @default.
- W3109771530 crossrefType "journal-article" @default.
- W3109771530 hasAuthorship W3109771530A5039442998 @default.
- W3109771530 hasAuthorship W3109771530A5074808403 @default.
- W3109771530 hasAuthorship W3109771530A5081679022 @default.
- W3109771530 hasBestOaLocation W31097715301 @default.
- W3109771530 hasConcept C115961682 @default.
- W3109771530 hasConcept C119857082 @default.
- W3109771530 hasConcept C132525143 @default.
- W3109771530 hasConcept C153180895 @default.
- W3109771530 hasConcept C154945302 @default.
- W3109771530 hasConcept C169760540 @default.
- W3109771530 hasConcept C179372163 @default.
- W3109771530 hasConcept C204321447 @default.
- W3109771530 hasConcept C205711294 @default.
- W3109771530 hasConcept C26760741 @default.
- W3109771530 hasConcept C2987255567 @default.
- W3109771530 hasConcept C41008148 @default.
- W3109771530 hasConcept C52146309 @default.
- W3109771530 hasConcept C75294576 @default.
- W3109771530 hasConcept C80444323 @default.
- W3109771530 hasConcept C86803240 @default.
- W3109771530 hasConceptScore W3109771530C115961682 @default.
- W3109771530 hasConceptScore W3109771530C119857082 @default.
- W3109771530 hasConceptScore W3109771530C132525143 @default.
- W3109771530 hasConceptScore W3109771530C153180895 @default.
- W3109771530 hasConceptScore W3109771530C154945302 @default.
- W3109771530 hasConceptScore W3109771530C169760540 @default.
- W3109771530 hasConceptScore W3109771530C179372163 @default.
- W3109771530 hasConceptScore W3109771530C204321447 @default.
- W3109771530 hasConceptScore W3109771530C205711294 @default.
- W3109771530 hasConceptScore W3109771530C26760741 @default.
- W3109771530 hasConceptScore W3109771530C2987255567 @default.