Matches in SemOpenAlex for { <https://semopenalex.org/work/W3011515866> ?p ?o ?g. }
- W3011515866 abstract "Several methods have been proposed to explain the decisions of neural networks in the visual domain via saliency heatmaps (aka relevances/feature importance scores). Thus far, these methods were mainly validated on real-world images, using either pixel perturbation experiments or bounding box localization accuracies. In the present work, we propose instead to evaluate explanations in a restricted and controlled setup using a synthetic dataset of rendered 3D shapes. To this end, we generate a CLEVR-alike visual question answering benchmark with around 40,000 questions, where the ground truth pixel coordinates of relevant objects are known, which allows us to validate explanations in a fair and transparent way. We further introduce two straightforward metrics to evaluate explanations in this setup, and compare their outcomes to standard pixel perturbation using a Relation Network model and three decomposition-based explanation methods: Gradient x Input, Integrated Gradients and Layer-wise Relevance Propagation. Among the tested methods, Layer-wise Relevance Propagation was shown to perform best, followed by Integrated Gradients. More generally, we expect the release of our dataset and code to support the development and comparison of methods on a well-defined common ground." @default.
- W3011515866 created "2020-03-23" @default.
- W3011515866 creator A5026451495 @default.
- W3011515866 creator A5035184440 @default.
- W3011515866 creator A5059359500 @default.
- W3011515866 date "2020-03-16" @default.
- W3011515866 modified "2023-10-16" @default.
- W3011515866 title "Towards Ground Truth Evaluation of Visual Explanations" @default.
- W3011515866 cites W1787224781 @default.
- W3011515866 cites W1849277567 @default.
- W3011515866 cites W1969404656 @default.
- W3011515866 cites W2083844448 @default.
- W3011515866 cites W2117539524 @default.
- W3011515866 cites W2195388612 @default.
- W3011515866 cites W2240067561 @default.
- W3011515866 cites W2282821441 @default.
- W3011515866 cites W2346578521 @default.
- W3011515866 cites W2503388974 @default.
- W3011515866 cites W2518775244 @default.
- W3011515866 cites W2527189750 @default.
- W3011515866 cites W2560730294 @default.
- W3011515866 cites W2561412020 @default.
- W3011515866 cites W2561715562 @default.
- W3011515866 cites W2581082771 @default.
- W3011515866 cites W2594633041 @default.
- W3011515866 cites W2626639386 @default.
- W3011515866 cites W2657631929 @default.
- W3011515866 cites W2776207810 @default.
- W3011515866 cites W2780718008 @default.
- W3011515866 cites W2785648721 @default.
- W3011515866 cites W2785760873 @default.
- W3011515866 cites W2794583223 @default.
- W3011515866 cites W2798698226 @default.
- W3011515866 cites W2806219033 @default.
- W3011515866 cites W2891612330 @default.
- W3011515866 cites W2903165775 @default.
- W3011515866 cites W2949725973 @default.
- W3011515866 cites W2950761309 @default.
- W3011515866 cites W2951152347 @default.
- W3011515866 cites W2962851944 @default.
- W3011515866 cites W2962862931 @default.
- W3011515866 cites W2963341956 @default.
- W3011515866 cites W2963349562 @default.
- W3011515866 cites W2963365341 @default.
- W3011515866 cites W2963382180 @default.
- W3011515866 cites W2963644680 @default.
- W3011515866 cites W2963907629 @default.
- W3011515866 cites W2970447476 @default.
- W3011515866 cites W2973136764 @default.
- W3011515866 cites W2982699810 @default.
- W3011515866 cites W2997428643 @default.
- W3011515866 cites W3006064320 @default.
- W3011515866 cites W3091589898 @default.
- W3011515866 cites W3101609372 @default.
- W3011515866 cites W3102564565 @default.
- W3011515866 cites W3105115233 @default.
- W3011515866 hasPublicationYear "2020" @default.
- W3011515866 type Work @default.
- W3011515866 sameAs 3011515866 @default.
- W3011515866 citedByCount "10" @default.
- W3011515866 countsByYear W30115158662020 @default.
- W3011515866 countsByYear W30115158662021 @default.
- W3011515866 countsByYear W30115158662023 @default.
- W3011515866 crossrefType "posted-content" @default.
- W3011515866 hasAuthorship W3011515866A5026451495 @default.
- W3011515866 hasAuthorship W3011515866A5035184440 @default.
- W3011515866 hasAuthorship W3011515866A5059359500 @default.
- W3011515866 hasConcept C11413529 @default.
- W3011515866 hasConcept C115961682 @default.
- W3011515866 hasConcept C119857082 @default.
- W3011515866 hasConcept C121158502 @default.
- W3011515866 hasConcept C146849305 @default.
- W3011515866 hasConcept C147037132 @default.
- W3011515866 hasConcept C153180895 @default.
- W3011515866 hasConcept C154945302 @default.
- W3011515866 hasConcept C158154518 @default.
- W3011515866 hasConcept C160633673 @default.
- W3011515866 hasConcept C161191863 @default.
- W3011515866 hasConcept C17744445 @default.
- W3011515866 hasConcept C185798385 @default.
- W3011515866 hasConcept C199539241 @default.
- W3011515866 hasConcept C205649164 @default.
- W3011515866 hasConcept C41008148 @default.
- W3011515866 hasConcept C50644808 @default.
- W3011515866 hasConcept C58640448 @default.
- W3011515866 hasConcept C63584917 @default.
- W3011515866 hasConceptScore W3011515866C11413529 @default.
- W3011515866 hasConceptScore W3011515866C115961682 @default.
- W3011515866 hasConceptScore W3011515866C119857082 @default.
- W3011515866 hasConceptScore W3011515866C121158502 @default.
- W3011515866 hasConceptScore W3011515866C146849305 @default.
- W3011515866 hasConceptScore W3011515866C147037132 @default.
- W3011515866 hasConceptScore W3011515866C153180895 @default.
- W3011515866 hasConceptScore W3011515866C154945302 @default.
- W3011515866 hasConceptScore W3011515866C158154518 @default.
- W3011515866 hasConceptScore W3011515866C160633673 @default.
- W3011515866 hasConceptScore W3011515866C161191863 @default.
- W3011515866 hasConceptScore W3011515866C17744445 @default.
- W3011515866 hasConceptScore W3011515866C185798385 @default.
- W3011515866 hasConceptScore W3011515866C199539241 @default.