Matches in SemOpenAlex for { <https://semopenalex.org/work/W3034854924> ?p ?o ?g. }
Showing items 1 to 89 of
89
with 100 items per page.
- W3034854924 abstract "Existing VQA datasets contain questions with varying levels of complexity. While the majority of questions in these datasets require perception for recognizing existence, properties, and spatial relationships of entities, a significant portion of questions pose challenges that correspond to reasoning tasks - tasks that can only be answered through a synthesis of perception and knowledge about the world, logic and / or reasoning. Analyzing performance across this distinction allows us to notice when existing VQA models have consistency issues - they answer the reasoning questions correctly but fail on associated low-level perception questions. For example, in Figure 1, models answer the complex reasoning question “Is the banana ripe enough to eat?” correctly, but fail on the associated perception question “Are the bananas mostly green or yellow?” indicating that the model likely answered the reasoning question correctly but for the wrong reason. We quantify the extent to which this phenomenon occurs by creating a new Reasoning split of the VQA dataset and collecting VQAintrospect, a new dataset1 which currently consists of 200K new perception questions which serve as sub questions corresponding to the set of perceptual tasks needed to effectively answer the complex reasoning questions in the Reasoning split. Our evaluation shows that state-of-the-art VQA models have comparable performance in answering perception and reasoning questions, but suffer from consistency problems. To address this shortcoming, we propose an approach called Sub-Question Importance-aware Network Tuning (SQuINT), which encourages the model to attend to the same parts of the image when answering the reasoning question and the perception sub question. We show that SQuINT improves model consistency by ~7%, also marginally improving performance on the Reasoning questions in VQA, while also displaying better attention maps." @default.
- W3034854924 created "2020-06-19" @default.
- W3034854924 creator A5011998621 @default.
- W3034854924 creator A5019726734 @default.
- W3034854924 creator A5028114802 @default.
- W3034854924 creator A5043228682 @default.
- W3034854924 creator A5046238088 @default.
- W3034854924 creator A5050342343 @default.
- W3034854924 creator A5059781275 @default.
- W3034854924 date "2020-06-01" @default.
- W3034854924 modified "2023-10-16" @default.
- W3034854924 title "SQuINTing at VQA Models: Introspecting VQA Models With Sub-Questions" @default.
- W3034854924 cites W1501418839 @default.
- W3034854924 cites W1682403713 @default.
- W3034854924 cites W1933349210 @default.
- W3034854924 cites W2118373646 @default.
- W3034854924 cites W2560730294 @default.
- W3034854924 cites W2759653627 @default.
- W3034854924 cites W2953039212 @default.
- W3034854924 cites W2962858109 @default.
- W3034854924 cites W2963518342 @default.
- W3034854924 cites W2963609017 @default.
- W3034854924 cites W2963644680 @default.
- W3034854924 cites W2963890019 @default.
- W3034854924 cites W2964061310 @default.
- W3034854924 cites W2965628639 @default.
- W3034854924 cites W2983256121 @default.
- W3034854924 doi "https://doi.org/10.1109/cvpr42600.2020.01002" @default.
- W3034854924 hasPublicationYear "2020" @default.
- W3034854924 type Work @default.
- W3034854924 sameAs 3034854924 @default.
- W3034854924 citedByCount "22" @default.
- W3034854924 countsByYear W30348549242020 @default.
- W3034854924 countsByYear W30348549242021 @default.
- W3034854924 countsByYear W30348549242022 @default.
- W3034854924 countsByYear W30348549242023 @default.
- W3034854924 crossrefType "proceedings-article" @default.
- W3034854924 hasAuthorship W3034854924A5011998621 @default.
- W3034854924 hasAuthorship W3034854924A5019726734 @default.
- W3034854924 hasAuthorship W3034854924A5028114802 @default.
- W3034854924 hasAuthorship W3034854924A5043228682 @default.
- W3034854924 hasAuthorship W3034854924A5046238088 @default.
- W3034854924 hasAuthorship W3034854924A5050342343 @default.
- W3034854924 hasAuthorship W3034854924A5059781275 @default.
- W3034854924 hasBestOaLocation W30348549242 @default.
- W3034854924 hasConcept C111472728 @default.
- W3034854924 hasConcept C138885662 @default.
- W3034854924 hasConcept C154945302 @default.
- W3034854924 hasConcept C177264268 @default.
- W3034854924 hasConcept C17744445 @default.
- W3034854924 hasConcept C199360897 @default.
- W3034854924 hasConcept C199539241 @default.
- W3034854924 hasConcept C23123220 @default.
- W3034854924 hasConcept C26760741 @default.
- W3034854924 hasConcept C2776436953 @default.
- W3034854924 hasConcept C2779913896 @default.
- W3034854924 hasConcept C41008148 @default.
- W3034854924 hasConcept C44291984 @default.
- W3034854924 hasConceptScore W3034854924C111472728 @default.
- W3034854924 hasConceptScore W3034854924C138885662 @default.
- W3034854924 hasConceptScore W3034854924C154945302 @default.
- W3034854924 hasConceptScore W3034854924C177264268 @default.
- W3034854924 hasConceptScore W3034854924C17744445 @default.
- W3034854924 hasConceptScore W3034854924C199360897 @default.
- W3034854924 hasConceptScore W3034854924C199539241 @default.
- W3034854924 hasConceptScore W3034854924C23123220 @default.
- W3034854924 hasConceptScore W3034854924C26760741 @default.
- W3034854924 hasConceptScore W3034854924C2776436953 @default.
- W3034854924 hasConceptScore W3034854924C2779913896 @default.
- W3034854924 hasConceptScore W3034854924C41008148 @default.
- W3034854924 hasConceptScore W3034854924C44291984 @default.
- W3034854924 hasLocation W30348549241 @default.
- W3034854924 hasLocation W30348549242 @default.
- W3034854924 hasOpenAccess W3034854924 @default.
- W3034854924 hasPrimaryLocation W30348549241 @default.
- W3034854924 hasRelatedWork W15319282 @default.
- W3034854924 hasRelatedWork W1594455022 @default.
- W3034854924 hasRelatedWork W2123793327 @default.
- W3034854924 hasRelatedWork W2296730655 @default.
- W3034854924 hasRelatedWork W2351286801 @default.
- W3034854924 hasRelatedWork W2356380379 @default.
- W3034854924 hasRelatedWork W2357241418 @default.
- W3034854924 hasRelatedWork W2361152157 @default.
- W3034854924 hasRelatedWork W2805599431 @default.
- W3034854924 hasRelatedWork W4255117927 @default.
- W3034854924 isParatext "false" @default.
- W3034854924 isRetracted "false" @default.
- W3034854924 magId "3034854924" @default.
- W3034854924 workType "article" @default.