Matches in SemOpenAlex for { <https://semopenalex.org/work/W4308244299> ?p ?o ?g. }
Showing items 1 to 76 of
76
with 100 items per page.
- W4308244299 abstract "There is limited understanding of the information captured by deep spatiotemporal models in their intermediate representations. For example, while evidence suggests that action recognition algorithms are heavily influenced by visual appearance in single frames, no quantitative methodology exists for evaluating such static bias in the latent representation compared to bias toward dynamics. We tackle this challenge by proposing an approach for quantifying the static and dynamic biases of any spatiotemporal model, and apply our approach to three tasks, action recognition, automatic video object segmentation (AVOS) and video instance segmentation (VIS). Our key findings are: (i) Most examined models are biased toward static information. (ii) Some datasets that are assumed to be biased toward dynamics are actually biased toward static information. (iii) Individual channels in an architecture can be biased toward static, dynamic or a combination of the two. (iv) Most models converge to their culminating biases in the first half of training. We then explore how these biases affect performance on dynamically biased datasets. For action recognition, we propose StaticDropout, a semantically guided dropout that debiases a model from static information toward dynamics. For AVOS, we design a better combination of fusion and cross connection layers compared with previous architectures." @default.
- W4308244299 created "2022-11-09" @default.
- W4308244299 creator A5022335103 @default.
- W4308244299 creator A5033145377 @default.
- W4308244299 creator A5052269532 @default.
- W4308244299 creator A5072076034 @default.
- W4308244299 creator A5087848827 @default.
- W4308244299 creator A5089232277 @default.
- W4308244299 date "2022-11-03" @default.
- W4308244299 modified "2023-09-27" @default.
- W4308244299 title "Quantifying and Learning Static vs. Dynamic Information in Deep Spatiotemporal Networks" @default.
- W4308244299 doi "https://doi.org/10.48550/arxiv.2211.01783" @default.
- W4308244299 hasPublicationYear "2022" @default.
- W4308244299 type Work @default.
- W4308244299 citedByCount "0" @default.
- W4308244299 crossrefType "posted-content" @default.
- W4308244299 hasAuthorship W4308244299A5022335103 @default.
- W4308244299 hasAuthorship W4308244299A5033145377 @default.
- W4308244299 hasAuthorship W4308244299A5052269532 @default.
- W4308244299 hasAuthorship W4308244299A5072076034 @default.
- W4308244299 hasAuthorship W4308244299A5087848827 @default.
- W4308244299 hasAuthorship W4308244299A5089232277 @default.
- W4308244299 hasBestOaLocation W43082442991 @default.
- W4308244299 hasConcept C119857082 @default.
- W4308244299 hasConcept C121332964 @default.
- W4308244299 hasConcept C145912823 @default.
- W4308244299 hasConcept C154945302 @default.
- W4308244299 hasConcept C17744445 @default.
- W4308244299 hasConcept C19768560 @default.
- W4308244299 hasConcept C199539241 @default.
- W4308244299 hasConcept C24890656 @default.
- W4308244299 hasConcept C26517878 @default.
- W4308244299 hasConcept C2776145597 @default.
- W4308244299 hasConcept C2776359362 @default.
- W4308244299 hasConcept C2780791683 @default.
- W4308244299 hasConcept C2781238097 @default.
- W4308244299 hasConcept C38652104 @default.
- W4308244299 hasConcept C41008148 @default.
- W4308244299 hasConcept C62520636 @default.
- W4308244299 hasConcept C89600930 @default.
- W4308244299 hasConcept C94625758 @default.
- W4308244299 hasConceptScore W4308244299C119857082 @default.
- W4308244299 hasConceptScore W4308244299C121332964 @default.
- W4308244299 hasConceptScore W4308244299C145912823 @default.
- W4308244299 hasConceptScore W4308244299C154945302 @default.
- W4308244299 hasConceptScore W4308244299C17744445 @default.
- W4308244299 hasConceptScore W4308244299C19768560 @default.
- W4308244299 hasConceptScore W4308244299C199539241 @default.
- W4308244299 hasConceptScore W4308244299C24890656 @default.
- W4308244299 hasConceptScore W4308244299C26517878 @default.
- W4308244299 hasConceptScore W4308244299C2776145597 @default.
- W4308244299 hasConceptScore W4308244299C2776359362 @default.
- W4308244299 hasConceptScore W4308244299C2780791683 @default.
- W4308244299 hasConceptScore W4308244299C2781238097 @default.
- W4308244299 hasConceptScore W4308244299C38652104 @default.
- W4308244299 hasConceptScore W4308244299C41008148 @default.
- W4308244299 hasConceptScore W4308244299C62520636 @default.
- W4308244299 hasConceptScore W4308244299C89600930 @default.
- W4308244299 hasConceptScore W4308244299C94625758 @default.
- W4308244299 hasLocation W43082442991 @default.
- W4308244299 hasLocation W43082442992 @default.
- W4308244299 hasOpenAccess W4308244299 @default.
- W4308244299 hasPrimaryLocation W43082442991 @default.
- W4308244299 hasRelatedWork W1549289070 @default.
- W4308244299 hasRelatedWork W1963494852 @default.
- W4308244299 hasRelatedWork W2004370856 @default.
- W4308244299 hasRelatedWork W2019566805 @default.
- W4308244299 hasRelatedWork W2076514466 @default.
- W4308244299 hasRelatedWork W2383464976 @default.
- W4308244299 hasRelatedWork W2968409168 @default.
- W4308244299 hasRelatedWork W4280592718 @default.
- W4308244299 hasRelatedWork W4319792277 @default.
- W4308244299 hasRelatedWork W1967061043 @default.
- W4308244299 isParatext "false" @default.
- W4308244299 isRetracted "false" @default.
- W4308244299 workType "article" @default.