Matches in SemOpenAlex for { <https://semopenalex.org/work/W3202223295> ?p ?o ?g. }
Showing items 1 to 83 of
83
with 100 items per page.
- W3202223295 abstract "The problem of long-tailed recognition, where the number of examples per class is highly unbalanced, is considered. It is hypothesized that the well known tendency of standard classifier training to overfit to popular classes can be exploited for effective transfer learning. Rather than eliminating this overfitting, e.g. by adopting popular class-balanced sampling methods, the learning algorithm should instead leverage this overfitting to transfer geometric information from popular to low-shot classes. A new classifier architecture, GistNet, is proposed to support this goal, using constellations of classifier parameters to encode the class geometry. A new learning algorithm is then proposed for GeometrIc Structure Transfer (GIST), with resort to a combination of loss functions that combine class-balanced and random sampling to guarantee that, while overfitting to the popular classes is restricted to geometric parameters, it is leveraged to transfer class geometry from popular to few-shot classes. This enables better generalization for few-shot classes without the need for the manual specification of class weights, or even the explicit grouping of classes into different types. Experiments on two popular long-tailed recognition datasets show that GistNet outperforms existing solutions to this problem." @default.
- W3202223295 created "2021-10-11" @default.
- W3202223295 creator A5043325212 @default.
- W3202223295 creator A5081114810 @default.
- W3202223295 creator A5081304526 @default.
- W3202223295 creator A5083511236 @default.
- W3202223295 creator A5090815103 @default.
- W3202223295 date "2021-10-01" @default.
- W3202223295 modified "2023-09-30" @default.
- W3202223295 title "GistNet: a Geometric Structure Transfer Network for Long-Tailed Recognition" @default.
- W3202223295 cites W2108598243 @default.
- W3202223295 cites W2194775991 @default.
- W3202223295 cites W2781292787 @default.
- W3202223295 cites W2797977484 @default.
- W3202223295 cites W2798836702 @default.
- W3202223295 cites W2962876782 @default.
- W3202223295 cites W2962933664 @default.
- W3202223295 cites W2963078860 @default.
- W3202223295 cites W2963351448 @default.
- W3202223295 cites W2963691377 @default.
- W3202223295 cites W2963845150 @default.
- W3202223295 cites W2964105864 @default.
- W3202223295 cites W3034369739 @default.
- W3202223295 cites W3034601242 @default.
- W3202223295 doi "https://doi.org/10.1109/iccv48922.2021.00810" @default.
- W3202223295 hasPublicationYear "2021" @default.
- W3202223295 type Work @default.
- W3202223295 sameAs 3202223295 @default.
- W3202223295 citedByCount "12" @default.
- W3202223295 countsByYear W32022232952021 @default.
- W3202223295 countsByYear W32022232952022 @default.
- W3202223295 countsByYear W32022232952023 @default.
- W3202223295 crossrefType "proceedings-article" @default.
- W3202223295 hasAuthorship W3202223295A5043325212 @default.
- W3202223295 hasAuthorship W3202223295A5081114810 @default.
- W3202223295 hasAuthorship W3202223295A5081304526 @default.
- W3202223295 hasAuthorship W3202223295A5083511236 @default.
- W3202223295 hasAuthorship W3202223295A5090815103 @default.
- W3202223295 hasBestOaLocation W32022232952 @default.
- W3202223295 hasConcept C11413529 @default.
- W3202223295 hasConcept C115961682 @default.
- W3202223295 hasConcept C119857082 @default.
- W3202223295 hasConcept C150899416 @default.
- W3202223295 hasConcept C153083717 @default.
- W3202223295 hasConcept C153180895 @default.
- W3202223295 hasConcept C154945302 @default.
- W3202223295 hasConcept C22019652 @default.
- W3202223295 hasConcept C2777212361 @default.
- W3202223295 hasConcept C41008148 @default.
- W3202223295 hasConcept C50644808 @default.
- W3202223295 hasConcept C75294576 @default.
- W3202223295 hasConcept C95623464 @default.
- W3202223295 hasConceptScore W3202223295C11413529 @default.
- W3202223295 hasConceptScore W3202223295C115961682 @default.
- W3202223295 hasConceptScore W3202223295C119857082 @default.
- W3202223295 hasConceptScore W3202223295C150899416 @default.
- W3202223295 hasConceptScore W3202223295C153083717 @default.
- W3202223295 hasConceptScore W3202223295C153180895 @default.
- W3202223295 hasConceptScore W3202223295C154945302 @default.
- W3202223295 hasConceptScore W3202223295C22019652 @default.
- W3202223295 hasConceptScore W3202223295C2777212361 @default.
- W3202223295 hasConceptScore W3202223295C41008148 @default.
- W3202223295 hasConceptScore W3202223295C50644808 @default.
- W3202223295 hasConceptScore W3202223295C75294576 @default.
- W3202223295 hasConceptScore W3202223295C95623464 @default.
- W3202223295 hasLocation W32022232951 @default.
- W3202223295 hasLocation W32022232952 @default.
- W3202223295 hasOpenAccess W3202223295 @default.
- W3202223295 hasPrimaryLocation W32022232951 @default.
- W3202223295 hasRelatedWork W2563096758 @default.
- W3202223295 hasRelatedWork W2742991909 @default.
- W3202223295 hasRelatedWork W2767651786 @default.
- W3202223295 hasRelatedWork W2972035100 @default.
- W3202223295 hasRelatedWork W2989932438 @default.
- W3202223295 hasRelatedWork W3012393889 @default.
- W3202223295 hasRelatedWork W3157856671 @default.
- W3202223295 hasRelatedWork W3195780658 @default.
- W3202223295 hasRelatedWork W4210691109 @default.
- W3202223295 hasRelatedWork W4220996320 @default.
- W3202223295 isParatext "false" @default.
- W3202223295 isRetracted "false" @default.
- W3202223295 magId "3202223295" @default.
- W3202223295 workType "article" @default.