Matches in SemOpenAlex for { <https://semopenalex.org/work/W4312593602> ?p ?o ?g. }
- W4312593602 endingPage "375" @default.
- W4312593602 startingPage "358" @default.
- W4312593602 abstract "The task of action detection aims at deducing both the action category and localization of the start and end moment for each action instance in a long, untrimmed video. While vision Transformers have driven the recent advances in video understanding, it is non-trivial to design an efficient architecture for action detection due to the prohibitively expensive self-attentions over a long sequence of video clips. To this end, we present an efficient hierarchical Spatio-Temporal Pyramid Transformer (STPT) for action detection, building upon the fact that the early self-attention layers in Transformers still focus on local patterns. Specifically, we propose to use local window attention to encode rich local spatio-temporal representations in the early stages while applying global attention modules to capture long-term space-time dependencies in the later stages. In this way, our STPT can encode both locality and dependency with largely reduced redundancy, delivering a promising trade-off between accuracy and efficiency. For example, with only RGB input, the proposed STPT achieves 53.6% mAP on THUMOS14, surpassing I3D+AFSD RGB model by over 10% and performing favorably against state-of-the-art AFSD that uses additional flow features with 31% fewer GFLOPs, which serves as an effective and efficient end-to-end Transformer-based framework for action detection. Code is available at https://github.com/ziplab/STPT ." @default.
- W4312593602 created "2023-01-05" @default.
- W4312593602 creator A5006461574 @default.
- W4312593602 creator A5018027213 @default.
- W4312593602 creator A5034967388 @default.
- W4312593602 creator A5066841335 @default.
- W4312593602 creator A5076928390 @default.
- W4312593602 date "2022-01-01" @default.
- W4312593602 modified "2023-10-03" @default.
- W4312593602 title "An Efficient Spatio-Temporal Pyramid Transformer for Action Detection" @default.
- W4312593602 cites W1522734439 @default.
- W4312593602 cites W1578985305 @default.
- W4312593602 cites W1927052826 @default.
- W4312593602 cites W2597958930 @default.
- W4312593602 cites W2755876276 @default.
- W4312593602 cites W2952435096 @default.
- W4312593602 cites W2962677524 @default.
- W4312593602 cites W2962876901 @default.
- W4312593602 cites W2963112696 @default.
- W4312593602 cites W2963247196 @default.
- W4312593602 cites W2963299740 @default.
- W4312593602 cites W2963315828 @default.
- W4312593602 cites W2963321993 @default.
- W4312593602 cites W2963351448 @default.
- W4312593602 cites W2963524571 @default.
- W4312593602 cites W2963820951 @default.
- W4312593602 cites W2964121718 @default.
- W4312593602 cites W2964216549 @default.
- W4312593602 cites W2983918066 @default.
- W4312593602 cites W2986407524 @default.
- W4312593602 cites W2990503944 @default.
- W4312593602 cites W2997314266 @default.
- W4312593602 cites W2998486508 @default.
- W4312593602 cites W3034623254 @default.
- W4312593602 cites W3035251589 @default.
- W4312593602 cites W3069380482 @default.
- W4312593602 cites W3100481960 @default.
- W4312593602 cites W3106041614 @default.
- W4312593602 cites W3128626728 @default.
- W4312593602 cites W3131500599 @default.
- W4312593602 cites W3138516171 @default.
- W4312593602 cites W3145586615 @default.
- W4312593602 cites W3170642968 @default.
- W4312593602 cites W3176444885 @default.
- W4312593602 cites W3202003978 @default.
- W4312593602 cites W3202076256 @default.
- W4312593602 cites W3210279979 @default.
- W4312593602 cites W4200630755 @default.
- W4312593602 cites W4200631626 @default.
- W4312593602 cites W4214493665 @default.
- W4312593602 cites W4214516465 @default.
- W4312593602 cites W4214612132 @default.
- W4312593602 cites W4214614183 @default.
- W4312593602 cites W4214633470 @default.
- W4312593602 cites W4214661601 @default.
- W4312593602 cites W4214736059 @default.
- W4312593602 cites W4226248135 @default.
- W4312593602 cites W4312919330 @default.
- W4312593602 doi "https://doi.org/10.1007/978-3-031-19830-4_21" @default.
- W4312593602 hasPublicationYear "2022" @default.
- W4312593602 type Work @default.
- W4312593602 citedByCount "4" @default.
- W4312593602 countsByYear W43125936022023 @default.
- W4312593602 crossrefType "book-chapter" @default.
- W4312593602 hasAuthorship W4312593602A5006461574 @default.
- W4312593602 hasAuthorship W4312593602A5018027213 @default.
- W4312593602 hasAuthorship W4312593602A5034967388 @default.
- W4312593602 hasAuthorship W4312593602A5066841335 @default.
- W4312593602 hasAuthorship W4312593602A5076928390 @default.
- W4312593602 hasBestOaLocation W43125936022 @default.
- W4312593602 hasConcept C104317684 @default.
- W4312593602 hasConcept C111919701 @default.
- W4312593602 hasConcept C115537543 @default.
- W4312593602 hasConcept C121332964 @default.
- W4312593602 hasConcept C138885662 @default.
- W4312593602 hasConcept C152124472 @default.
- W4312593602 hasConcept C153180895 @default.
- W4312593602 hasConcept C154945302 @default.
- W4312593602 hasConcept C165801399 @default.
- W4312593602 hasConcept C173608175 @default.
- W4312593602 hasConcept C185592680 @default.
- W4312593602 hasConcept C27602214 @default.
- W4312593602 hasConcept C2779808786 @default.
- W4312593602 hasConcept C31972630 @default.
- W4312593602 hasConcept C41008148 @default.
- W4312593602 hasConcept C41895202 @default.
- W4312593602 hasConcept C55493867 @default.
- W4312593602 hasConcept C62520636 @default.
- W4312593602 hasConcept C66322947 @default.
- W4312593602 hasConcept C66746571 @default.
- W4312593602 hasConcept C82990744 @default.
- W4312593602 hasConceptScore W4312593602C104317684 @default.
- W4312593602 hasConceptScore W4312593602C111919701 @default.
- W4312593602 hasConceptScore W4312593602C115537543 @default.
- W4312593602 hasConceptScore W4312593602C121332964 @default.
- W4312593602 hasConceptScore W4312593602C138885662 @default.
- W4312593602 hasConceptScore W4312593602C152124472 @default.
- W4312593602 hasConceptScore W4312593602C153180895 @default.