Matches in SemOpenAlex for { <https://semopenalex.org/work/W4386072392> ?p ?o ?g. }
Showing items 1 to 82 of
82
with 100 items per page.
- W4386072392 abstract "Previous knowledge distillation methods have shown their impressive performance on model compression tasks, however, it is hard to explain how the knowledge they transferred helps to improve the performance of the student network. In this work, we focus on proposing a knowledge distillation method that has both high interpretability and competitive performance. We first revisit the structure of mainstream CNN models and reveal that possessing the capacity of identifying class discriminative regions of input is critical for CNN to perform classification. Furthermore, we demonstrate that this capacity can be obtained and enhanced by transferring class activation maps. Based on our findings, we propose class attention transfer based knowledge distillation (CAT-KD). Different from previous KD methods, we explore and present several properties of the knowledge transferred by our method, which not only improve the interpretability of CAT-KD but also contribute to a better understanding of CNN. While having high interpretability, CAT-KD achieves state-of-the-art performance on multiple benchmarks. Code is available at: https://github.com/GzyAftermath/CAT-KD." @default.
- W4386072392 created "2023-08-23" @default.
- W4386072392 creator A5016590465 @default.
- W4386072392 creator A5048417676 @default.
- W4386072392 creator A5060900881 @default.
- W4386072392 creator A5083787204 @default.
- W4386072392 date "2023-06-01" @default.
- W4386072392 modified "2023-10-11" @default.
- W4386072392 title "Class Attention Transfer Based Knowledge Distillation" @default.
- W4386072392 cites W2108598243 @default.
- W4386072392 cites W2295107390 @default.
- W4386072392 cites W2739879705 @default.
- W4386072392 cites W2765793020 @default.
- W4386072392 cites W2957702778 @default.
- W4386072392 cites W2963140444 @default.
- W4386072392 cites W2963606198 @default.
- W4386072392 cites W2963949812 @default.
- W4386072392 cites W2982157312 @default.
- W4386072392 cites W2982242214 @default.
- W4386072392 cites W2986015886 @default.
- W4386072392 cites W2997006708 @default.
- W4386072392 cites W3004127093 @default.
- W4386072392 cites W3024127982 @default.
- W4386072392 cites W3035253074 @default.
- W4386072392 cites W3177196641 @default.
- W4386072392 cites W3198054860 @default.
- W4386072392 cites W4226426325 @default.
- W4386072392 doi "https://doi.org/10.1109/cvpr52729.2023.01142" @default.
- W4386072392 hasPublicationYear "2023" @default.
- W4386072392 type Work @default.
- W4386072392 citedByCount "1" @default.
- W4386072392 crossrefType "proceedings-article" @default.
- W4386072392 hasAuthorship W4386072392A5016590465 @default.
- W4386072392 hasAuthorship W4386072392A5048417676 @default.
- W4386072392 hasAuthorship W4386072392A5060900881 @default.
- W4386072392 hasAuthorship W4386072392A5083787204 @default.
- W4386072392 hasConcept C119857082 @default.
- W4386072392 hasConcept C154945302 @default.
- W4386072392 hasConcept C177264268 @default.
- W4386072392 hasConcept C178790620 @default.
- W4386072392 hasConcept C185592680 @default.
- W4386072392 hasConcept C199360897 @default.
- W4386072392 hasConcept C204030448 @default.
- W4386072392 hasConcept C2776760102 @default.
- W4386072392 hasConcept C2776960227 @default.
- W4386072392 hasConcept C2777212361 @default.
- W4386072392 hasConcept C2781067378 @default.
- W4386072392 hasConcept C41008148 @default.
- W4386072392 hasConcept C56739046 @default.
- W4386072392 hasConcept C97931131 @default.
- W4386072392 hasConceptScore W4386072392C119857082 @default.
- W4386072392 hasConceptScore W4386072392C154945302 @default.
- W4386072392 hasConceptScore W4386072392C177264268 @default.
- W4386072392 hasConceptScore W4386072392C178790620 @default.
- W4386072392 hasConceptScore W4386072392C185592680 @default.
- W4386072392 hasConceptScore W4386072392C199360897 @default.
- W4386072392 hasConceptScore W4386072392C204030448 @default.
- W4386072392 hasConceptScore W4386072392C2776760102 @default.
- W4386072392 hasConceptScore W4386072392C2776960227 @default.
- W4386072392 hasConceptScore W4386072392C2777212361 @default.
- W4386072392 hasConceptScore W4386072392C2781067378 @default.
- W4386072392 hasConceptScore W4386072392C41008148 @default.
- W4386072392 hasConceptScore W4386072392C56739046 @default.
- W4386072392 hasConceptScore W4386072392C97931131 @default.
- W4386072392 hasFunder F4320310921 @default.
- W4386072392 hasFunder F4320321001 @default.
- W4386072392 hasLocation W43860723921 @default.
- W4386072392 hasOpenAccess W4386072392 @default.
- W4386072392 hasPrimaryLocation W43860723921 @default.
- W4386072392 hasRelatedWork W3006943036 @default.
- W4386072392 hasRelatedWork W3090617741 @default.
- W4386072392 hasRelatedWork W4200511449 @default.
- W4386072392 hasRelatedWork W4206534706 @default.
- W4386072392 hasRelatedWork W4229079080 @default.
- W4386072392 hasRelatedWork W4280605930 @default.
- W4386072392 hasRelatedWork W4299487748 @default.
- W4386072392 hasRelatedWork W4385957992 @default.
- W4386072392 hasRelatedWork W4385965371 @default.
- W4386072392 hasRelatedWork W4386025632 @default.
- W4386072392 isParatext "false" @default.
- W4386072392 isRetracted "false" @default.
- W4386072392 workType "article" @default.