FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3175466730> ?p ?o ?g. }

• W3175466730 abstract "This paper investigates two techniques for developing efficient self-supervised vision transformers (EsViT) for visual representation learning. First, we show through a comprehensive empirical study that multi-stage architectures with sparse self-attentions can significantly reduce modeling complexity but with a cost of losing the ability to capture fine-grained correspondences between image regions. Second, we propose a new pre-training task of region matching which allows the model to capture fine-grained region dependencies and as a result significantly improves the quality of the learned vision representations. Our results show that combining the two techniques, EsViT achieves 81.3% top-1 on the ImageNet linear probe evaluation, outperforming prior arts with around an order magnitude of higher throughput. When transferring to downstream linear classification tasks, EsViT outperforms its supervised counterpart on 17 out of 18 datasets. The code and models are publicly available: https://github.com/microsoft/esvit" @default.
• W3175466730 created "2021-07-05" @default.
• W3175466730 creator A5003075563 @default.
• W3175466730 creator A5030624782 @default.
• W3175466730 creator A5040938049 @default.
• W3175466730 creator A5046214153 @default.
• W3175466730 creator A5047233371 @default.
• W3175466730 creator A5057293861 @default.
• W3175466730 creator A5059735251 @default.
• W3175466730 creator A5081369301 @default.
• W3175466730 date "2021-06-17" @default.
• W3175466730 modified "2023-10-14" @default.
• W3175466730 title "Efficient Self-supervised Vision Transformers for Representation Learning" @default.
• W3175466730 cites W1686810756 @default.
• W3175466730 cites W2108598243 @default.
• W3175466730 cites W2144796873 @default.
• W3175466730 cites W2194775991 @default.
• W3175466730 cites W2308529009 @default.
• W3175466730 cites W2321533354 @default.
• W3175466730 cites W2326925005 @default.
• W3175466730 cites W2337374958 @default.
• W3175466730 cites W2342877626 @default.
• W3175466730 cites W2558661413 @default.
• W3175466730 cites W2622263826 @default.
• W3175466730 cites W2626778328 @default.
• W3175466730 cites W2743200750 @default.
• W3175466730 cites W2768282280 @default.
• W3175466730 cites W2842511635 @default.
• W3175466730 cites W2883725317 @default.
• W3175466730 cites W2896457183 @default.
• W3175466730 cites W2922509574 @default.
• W3175466730 cites W2948433173 @default.
• W3175466730 cites W2951326654 @default.
• W3175466730 cites W2951873722 @default.
• W3175466730 cites W2962742544 @default.
• W3175466730 cites W2964074409 @default.
• W3175466730 cites W2968124245 @default.
• W3175466730 cites W2968880719 @default.
• W3175466730 cites W2969862959 @default.
• W3175466730 cites W2969876226 @default.
• W3175466730 cites W2970241862 @default.
• W3175466730 cites W2970608575 @default.
• W3175466730 cites W2971155163 @default.
• W3175466730 cites W2975501350 @default.
• W3175466730 cites W2990500698 @default.
• W3175466730 cites W2997591391 @default.
• W3175466730 cites W3005680577 @default.
• W3175466730 cites W3007700590 @default.
• W3175466730 cites W3009561768 @default.
• W3175466730 cites W3022061250 @default.
• W3175466730 cites W3030163527 @default.
• W3175466730 cites W3034445277 @default.
• W3175466730 cites W3034576826 @default.
• W3175466730 cites W3035022492 @default.
• W3175466730 cites W3035060554 @default.
• W3175466730 cites W3035524453 @default.
• W3175466730 cites W3036224891 @default.
• W3175466730 cites W3036982689 @default.
• W3175466730 cites W3092462694 @default.
• W3175466730 cites W3096609285 @default.
• W3175466730 cites W3101415077 @default.
• W3175466730 cites W3103465009 @default.
• W3175466730 cites W3105236818 @default.
• W3175466730 cites W3106428938 @default.
• W3175466730 cites W3108995912 @default.
• W3175466730 cites W3109319753 @default.
• W3175466730 cites W3113997557 @default.
• W3175466730 cites W3116489684 @default.
• W3175466730 cites W3119786062 @default.
• W3175466730 cites W3121480429 @default.
• W3175466730 cites W3131500599 @default.
• W3175466730 cites W3135367836 @default.
• W3175466730 cites W3135715136 @default.
• W3175466730 cites W3138516171 @default.
• W3175466730 cites W3139633126 @default.
• W3175466730 cites W3139773203 @default.
• W3175466730 cites W3145450063 @default.
• W3175466730 cites W3157506437 @default.
• W3175466730 cites W3159481202 @default.
• W3175466730 cites W3160566314 @default.
• W3175466730 cites W3161438416 @default.
• W3175466730 cites W3168649818 @default.
• W3175466730 cites W3168822201 @default.
• W3175466730 cites W3179869055 @default.
• W3175466730 cites W343636949 @default.
• W3175466730 doi "https://doi.org/10.48550/arxiv.2106.09785" @default.
• W3175466730 hasPublicationYear "2021" @default.
• W3175466730 type Work @default.
• W3175466730 sameAs 3175466730 @default.
• W3175466730 citedByCount "8" @default.
• W3175466730 countsByYear W31754667302020 @default.
• W3175466730 countsByYear W31754667302021 @default.
• W3175466730 countsByYear W31754667302022 @default.
• W3175466730 crossrefType "posted-content" @default.
• W3175466730 hasAuthorship W3175466730A5003075563 @default.
• W3175466730 hasAuthorship W3175466730A5030624782 @default.
• W3175466730 hasAuthorship W3175466730A5040938049 @default.
• W3175466730 hasAuthorship W3175466730A5046214153 @default.
• W3175466730 hasAuthorship W3175466730A5047233371 @default.
• W3175466730 hasAuthorship W3175466730A5057293861 @default.

• next