Matches in SemOpenAlex for { <https://semopenalex.org/work/W4312882717> ?p ?o ?g. }
- W4312882717 endingPage "540" @default.
- W4312882717 startingPage "524" @default.
- W4312882717 abstract "This paper presents a transformer framework for few-shot learning, termed TransVLAD, with one focus showing the power of locally aggregated descriptors for few-shot learning. Our TransVLAD model is simple: a standard transformer encoder following a NeXtVLAD aggregation module to output the locally aggregated descriptors. In contrast to the prevailing use of CNN as part of the feature extractor, we are the first to prove self-supervised learning like masked autoencoders (MAE) can deal with the overfitting of transformers in few-shot image classification. Besides, few-shot learning can benefit from this general-purpose pre-training. Then, we propose two methods to mitigate few-shot biases, supervision bias and simple-characteristic bias. The first method is introducing masking operation into fine-tuning, by which we accelerate fine-tuning (by more than 3x) and improve accuracy. The second one is adapting focal loss into soft focal loss to focus on hard characteristics learning. Our TransVLAD finally tops 10 benchmarks on five popular few-shot datasets by an average of more than 2%." @default.
- W4312882717 created "2023-01-05" @default.
- W4312882717 creator A5030626654 @default.
- W4312882717 creator A5031648383 @default.
- W4312882717 creator A5063818643 @default.
- W4312882717 creator A5075967266 @default.
- W4312882717 creator A5076720384 @default.
- W4312882717 creator A5078021360 @default.
- W4312882717 date "2022-01-01" @default.
- W4312882717 modified "2023-09-27" @default.
- W4312882717 title "TransVLAD: Focusing on Locally Aggregated Descriptors for Few-Shot Learning" @default.
- W4312882717 cites W2012592962 @default.
- W4312882717 cites W2117539524 @default.
- W4312882717 cites W2165698076 @default.
- W4312882717 cites W2166742463 @default.
- W4312882717 cites W2786808285 @default.
- W4312882717 cites W2894737833 @default.
- W4312882717 cites W2900560726 @default.
- W4312882717 cites W2963070905 @default.
- W4312882717 cites W2963351448 @default.
- W4312882717 cites W2963845150 @default.
- W4312882717 cites W2964105864 @default.
- W4312882717 cites W2979689312 @default.
- W4312882717 cites W2994633389 @default.
- W4312882717 cites W3009081299 @default.
- W4312882717 cites W3012255272 @default.
- W4312882717 cites W3034587791 @default.
- W4312882717 cites W3035524453 @default.
- W4312882717 cites W3096609285 @default.
- W4312882717 cites W3108975329 @default.
- W4312882717 cites W3145450063 @default.
- W4312882717 cites W3159481202 @default.
- W4312882717 cites W3182874523 @default.
- W4312882717 cites W3185341429 @default.
- W4312882717 cites W3205249428 @default.
- W4312882717 cites W4214562728 @default.
- W4312882717 cites W4285287412 @default.
- W4312882717 cites W4312312750 @default.
- W4312882717 cites W4312349930 @default.
- W4312882717 cites W4313156423 @default.
- W4312882717 doi "https://doi.org/10.1007/978-3-031-20044-1_30" @default.
- W4312882717 hasPublicationYear "2022" @default.
- W4312882717 type Work @default.
- W4312882717 citedByCount "1" @default.
- W4312882717 countsByYear W43128827172023 @default.
- W4312882717 crossrefType "book-chapter" @default.
- W4312882717 hasAuthorship W4312882717A5030626654 @default.
- W4312882717 hasAuthorship W4312882717A5031648383 @default.
- W4312882717 hasAuthorship W4312882717A5063818643 @default.
- W4312882717 hasAuthorship W4312882717A5075967266 @default.
- W4312882717 hasAuthorship W4312882717A5076720384 @default.
- W4312882717 hasAuthorship W4312882717A5078021360 @default.
- W4312882717 hasConcept C111919701 @default.
- W4312882717 hasConcept C117978034 @default.
- W4312882717 hasConcept C118505674 @default.
- W4312882717 hasConcept C119857082 @default.
- W4312882717 hasConcept C120665830 @default.
- W4312882717 hasConcept C121332964 @default.
- W4312882717 hasConcept C127413603 @default.
- W4312882717 hasConcept C153180895 @default.
- W4312882717 hasConcept C154945302 @default.
- W4312882717 hasConcept C165801399 @default.
- W4312882717 hasConcept C192209626 @default.
- W4312882717 hasConcept C21880701 @default.
- W4312882717 hasConcept C22019652 @default.
- W4312882717 hasConcept C3019835501 @default.
- W4312882717 hasConcept C41008148 @default.
- W4312882717 hasConcept C50644808 @default.
- W4312882717 hasConcept C62520636 @default.
- W4312882717 hasConcept C66322947 @default.
- W4312882717 hasConceptScore W4312882717C111919701 @default.
- W4312882717 hasConceptScore W4312882717C117978034 @default.
- W4312882717 hasConceptScore W4312882717C118505674 @default.
- W4312882717 hasConceptScore W4312882717C119857082 @default.
- W4312882717 hasConceptScore W4312882717C120665830 @default.
- W4312882717 hasConceptScore W4312882717C121332964 @default.
- W4312882717 hasConceptScore W4312882717C127413603 @default.
- W4312882717 hasConceptScore W4312882717C153180895 @default.
- W4312882717 hasConceptScore W4312882717C154945302 @default.
- W4312882717 hasConceptScore W4312882717C165801399 @default.
- W4312882717 hasConceptScore W4312882717C192209626 @default.
- W4312882717 hasConceptScore W4312882717C21880701 @default.
- W4312882717 hasConceptScore W4312882717C22019652 @default.
- W4312882717 hasConceptScore W4312882717C3019835501 @default.
- W4312882717 hasConceptScore W4312882717C41008148 @default.
- W4312882717 hasConceptScore W4312882717C50644808 @default.
- W4312882717 hasConceptScore W4312882717C62520636 @default.
- W4312882717 hasConceptScore W4312882717C66322947 @default.
- W4312882717 hasLocation W43128827171 @default.
- W4312882717 hasOpenAccess W4312882717 @default.
- W4312882717 hasPrimaryLocation W43128827171 @default.
- W4312882717 hasRelatedWork W1996541855 @default.
- W4312882717 hasRelatedWork W2742991909 @default.
- W4312882717 hasRelatedWork W2767651786 @default.
- W4312882717 hasRelatedWork W2940336242 @default.
- W4312882717 hasRelatedWork W2989932438 @default.
- W4312882717 hasRelatedWork W3099765033 @default.
- W4312882717 hasRelatedWork W4210794429 @default.