FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W4313176063> ?p ?o ?g. }

• W4313176063 abstract "While real world challenges typically define visual categories with language words or phrases, most visual classification methods define categories with numerical indices. However, the language specification of the classes provides an especially useful prior for biased and noisy datasets, where it can help disambiguate what features are task-relevant. Recently, large-scale multimodal models have been shown to recognize a wide variety of high-level concepts from a language specification even without additional image training data, but they are often unable to distinguish classes for more fine-grained tasks. CNNs, in contrast, can extract subtle image features that are required for fine-grained discrimination, but will overfit to any bias or noise in datasets. Our insight is to use high-level language specification as advice for constraining the classification evidence to task-relevant features, instead of distractors. To do this, we ground task-relevant words or phrases with attention maps from a pretrained large-scale model. We then use this grounding to supervise a classifier's spatial attention away from distracting context. We show that supervising spatial attention in this way improves performance on classification tasks with biased and noisy data, including ~3 −15% worst-group accuracy improvements and ~41-45% relative improvements on fairness metrics." @default.
• W4313176063 created "2023-01-06" @default.
• W4313176063 creator A5020311381 @default.
• W4313176063 creator A5029105520 @default.
• W4313176063 creator A5034289489 @default.
• W4313176063 creator A5037747070 @default.
• W4313176063 creator A5055722778 @default.
• W4313176063 creator A5072427753 @default.
• W4313176063 date "2022-06-01" @default.
• W4313176063 modified "2023-09-27" @default.
• W4313176063 title "On Guiding Visual Attention with Language Specification" @default.
• W4313176063 cites W1773149199 @default.
• W4313176063 cites W2128532956 @default.
• W4313176063 cites W2145215286 @default.
• W4313176063 cites W2146950091 @default.
• W4313176063 cites W2194775991 @default.
• W4313176063 cites W2250384498 @default.
• W4313176063 cites W2251512949 @default.
• W4313176063 cites W2732026016 @default.
• W4313176063 cites W2776207810 @default.
• W4313176063 cites W2886641317 @default.
• W4313176063 cites W2962714319 @default.
• W4313176063 cites W2962735233 @default.
• W4313176063 cites W2962787423 @default.
• W4313176063 cites W2962858109 @default.
• W4313176063 cites W2962978560 @default.
• W4313176063 cites W2963350032 @default.
• W4313176063 cites W2963630207 @default.
• W4313176063 cites W2963811535 @default.
• W4313176063 cites W2963955958 @default.
• W4313176063 cites W2990751682 @default.
• W4313176063 cites W3013629728 @default.
• W4313176063 cites W3034410824 @default.
• W4313176063 cites W3035337382 @default.
• W4313176063 cites W3100895823 @default.
• W4313176063 cites W3102259594 @default.
• W4313176063 cites W3109985940 @default.
• W4313176063 cites W3120895209 @default.
• W4313176063 cites W4214777488 @default.
• W4313176063 doi "https://doi.org/10.1109/cvpr52688.2022.01756" @default.
• W4313176063 hasPublicationYear "2022" @default.
• W4313176063 type Work @default.
• W4313176063 citedByCount "1" @default.
• W4313176063 countsByYear W43131760632023 @default.
• W4313176063 crossrefType "proceedings-article" @default.
• W4313176063 hasAuthorship W4313176063A5020311381 @default.
• W4313176063 hasAuthorship W4313176063A5029105520 @default.
• W4313176063 hasAuthorship W4313176063A5034289489 @default.
• W4313176063 hasAuthorship W4313176063A5037747070 @default.
• W4313176063 hasAuthorship W4313176063A5055722778 @default.
• W4313176063 hasAuthorship W4313176063A5072427753 @default.
• W4313176063 hasBestOaLocation W43131760632 @default.
• W4313176063 hasConcept C119857082 @default.
• W4313176063 hasConcept C136197465 @default.
• W4313176063 hasConcept C137293760 @default.
• W4313176063 hasConcept C151730666 @default.
• W4313176063 hasConcept C154945302 @default.
• W4313176063 hasConcept C162324750 @default.
• W4313176063 hasConcept C175154964 @default.
• W4313176063 hasConcept C187736073 @default.
• W4313176063 hasConcept C204321447 @default.
• W4313176063 hasConcept C22019652 @default.
• W4313176063 hasConcept C2779343474 @default.
• W4313176063 hasConcept C2780451532 @default.
• W4313176063 hasConcept C41008148 @default.
• W4313176063 hasConcept C50644808 @default.
• W4313176063 hasConcept C86803240 @default.
• W4313176063 hasConcept C95623464 @default.
• W4313176063 hasConceptScore W4313176063C119857082 @default.
• W4313176063 hasConceptScore W4313176063C136197465 @default.
• W4313176063 hasConceptScore W4313176063C137293760 @default.
• W4313176063 hasConceptScore W4313176063C151730666 @default.
• W4313176063 hasConceptScore W4313176063C154945302 @default.
• W4313176063 hasConceptScore W4313176063C162324750 @default.
• W4313176063 hasConceptScore W4313176063C175154964 @default.
• W4313176063 hasConceptScore W4313176063C187736073 @default.
• W4313176063 hasConceptScore W4313176063C204321447 @default.
• W4313176063 hasConceptScore W4313176063C22019652 @default.
• W4313176063 hasConceptScore W4313176063C2779343474 @default.
• W4313176063 hasConceptScore W4313176063C2780451532 @default.
• W4313176063 hasConceptScore W4313176063C41008148 @default.
• W4313176063 hasConceptScore W4313176063C50644808 @default.
• W4313176063 hasConceptScore W4313176063C86803240 @default.
• W4313176063 hasConceptScore W4313176063C95623464 @default.
• W4313176063 hasFunder F4320307102 @default.
• W4313176063 hasFunder F4320307764 @default.
• W4313176063 hasFunder F4320308737 @default.
• W4313176063 hasFunder F4320309327 @default.
• W4313176063 hasFunder F4320310598 @default.
• W4313176063 hasLocation W43131760631 @default.
• W4313176063 hasLocation W43131760632 @default.
• W4313176063 hasOpenAccess W4313176063 @default.
• W4313176063 hasPrimaryLocation W43131760631 @default.
• W4313176063 hasRelatedWork W1996541855 @default.
• W4313176063 hasRelatedWork W2742991909 @default.
• W4313176063 hasRelatedWork W2940336242 @default.
• W4313176063 hasRelatedWork W2953328427 @default.
• W4313176063 hasRelatedWork W2989932438 @default.
• W4313176063 hasRelatedWork W3092365268 @default.
• W4313176063 hasRelatedWork W3099765033 @default.

• next