Matches in SemOpenAlex for { <https://semopenalex.org/work/W4385768003> ?p ?o ?g. }
Showing items 1 to 72 of
72
with 100 items per page.
- W4385768003 abstract "Knowledge distillation is an effective method for model compression. However, it is still a challenging topic to apply knowledge distillation to detection tasks. There are two key points resulting in poor distillation performance for detection tasks. One is the serious imbalance between foreground and background features, another one is that small object lacks enough feature representation. To solve the above issues, we propose a new distillation method named dual relation knowledge distillation (DRKD), including pixel-wise relation distillation and instance-wise relation distillation. The pixel-wise relation distillation embeds pixel-wise features in the graph space and applies graph convolution to capture the global pixel relation. By distilling the global pixel relation, the student detector can learn the relation between foreground and background features, and avoid the difficulty of distilling features directly for the feature imbalance issue. Besides, we find that instance-wise relation supplements valuable knowledge beyond independent features for small objects. Thus, the instance-wise relation distillation is designed, which calculates the similarity of different instances to obtain a relation matrix. More importantly, a relation filter module is designed to highlight valuable instance relations. The proposed dual relation knowledge distillation is general and can be easily applied for both one-stage and two-stage detectors. Our method achieves state-of-the-art performance, which improves Faster R-CNN based on ResNet50 from 38.4% to 41.6% mAP and improves RetinaNet based on ResNet50 from 37.4% to 40.3% mAP on COCO 2017." @default.
- W4385768003 created "2023-08-12" @default.
- W4385768003 creator A5015563980 @default.
- W4385768003 creator A5028517540 @default.
- W4385768003 creator A5069765960 @default.
- W4385768003 creator A5075717206 @default.
- W4385768003 date "2023-08-01" @default.
- W4385768003 modified "2023-09-28" @default.
- W4385768003 title "Dual Relation Knowledge Distillation for Object Detection" @default.
- W4385768003 doi "https://doi.org/10.24963/ijcai.2023/142" @default.
- W4385768003 hasPublicationYear "2023" @default.
- W4385768003 type Work @default.
- W4385768003 citedByCount "0" @default.
- W4385768003 crossrefType "proceedings-article" @default.
- W4385768003 hasAuthorship W4385768003A5015563980 @default.
- W4385768003 hasAuthorship W4385768003A5028517540 @default.
- W4385768003 hasAuthorship W4385768003A5069765960 @default.
- W4385768003 hasAuthorship W4385768003A5075717206 @default.
- W4385768003 hasBestOaLocation W43857680031 @default.
- W4385768003 hasConcept C119857082 @default.
- W4385768003 hasConcept C124101348 @default.
- W4385768003 hasConcept C132525143 @default.
- W4385768003 hasConcept C138885662 @default.
- W4385768003 hasConcept C153180895 @default.
- W4385768003 hasConcept C154945302 @default.
- W4385768003 hasConcept C160633673 @default.
- W4385768003 hasConcept C178790620 @default.
- W4385768003 hasConcept C185592680 @default.
- W4385768003 hasConcept C204030448 @default.
- W4385768003 hasConcept C25343380 @default.
- W4385768003 hasConcept C2776401178 @default.
- W4385768003 hasConcept C31972630 @default.
- W4385768003 hasConcept C41008148 @default.
- W4385768003 hasConcept C41895202 @default.
- W4385768003 hasConcept C45347329 @default.
- W4385768003 hasConcept C50644808 @default.
- W4385768003 hasConcept C80444323 @default.
- W4385768003 hasConceptScore W4385768003C119857082 @default.
- W4385768003 hasConceptScore W4385768003C124101348 @default.
- W4385768003 hasConceptScore W4385768003C132525143 @default.
- W4385768003 hasConceptScore W4385768003C138885662 @default.
- W4385768003 hasConceptScore W4385768003C153180895 @default.
- W4385768003 hasConceptScore W4385768003C154945302 @default.
- W4385768003 hasConceptScore W4385768003C160633673 @default.
- W4385768003 hasConceptScore W4385768003C178790620 @default.
- W4385768003 hasConceptScore W4385768003C185592680 @default.
- W4385768003 hasConceptScore W4385768003C204030448 @default.
- W4385768003 hasConceptScore W4385768003C25343380 @default.
- W4385768003 hasConceptScore W4385768003C2776401178 @default.
- W4385768003 hasConceptScore W4385768003C31972630 @default.
- W4385768003 hasConceptScore W4385768003C41008148 @default.
- W4385768003 hasConceptScore W4385768003C41895202 @default.
- W4385768003 hasConceptScore W4385768003C45347329 @default.
- W4385768003 hasConceptScore W4385768003C50644808 @default.
- W4385768003 hasConceptScore W4385768003C80444323 @default.
- W4385768003 hasLocation W43857680031 @default.
- W4385768003 hasLocation W43857680032 @default.
- W4385768003 hasOpenAccess W4385768003 @default.
- W4385768003 hasPrimaryLocation W43857680031 @default.
- W4385768003 hasRelatedWork W121273120 @default.
- W4385768003 hasRelatedWork W2002009170 @default.
- W4385768003 hasRelatedWork W2034462085 @default.
- W4385768003 hasRelatedWork W2136485282 @default.
- W4385768003 hasRelatedWork W2141888456 @default.
- W4385768003 hasRelatedWork W2337415362 @default.
- W4385768003 hasRelatedWork W2546871836 @default.
- W4385768003 hasRelatedWork W2740820121 @default.
- W4385768003 hasRelatedWork W317572212 @default.
- W4385768003 hasRelatedWork W4312857205 @default.
- W4385768003 isParatext "false" @default.
- W4385768003 isRetracted "false" @default.
- W4385768003 workType "article" @default.