Matches in SemOpenAlex for { <https://semopenalex.org/work/W4387678029> ?p ?o ?g. }
Showing items 1 to 94 of
94
with 100 items per page.
- W4387678029 abstract "In medical image segmentation, the greatly successful U-Net structure has become the industry standard and has achieved great success. However, existing methods all directly fuse high-dimensional features with low-dimensional ones without considering the difference between high-dimensional and low-dimensional features, making it difficult to fit in small data scenarios and hence limiting the performance. This paper proposes a transformer-based method for feature fusion. Specifically, anchor attention and self-attention are used to extract features. The anchor attention effectively fuses high-dimensional and low-dimensional features, while self-attention extracts general features. Fusion of the high-dimensional and low-dimensional features solves the semantic gap between the two features. At the same time, the fusion model draws on comparative learning to guide the skipped connection and upsampling for the fusion. In particular, this Cross-Scale fusion method uses a downsampled feature embeded as a negative sample, and uses the skipped connection output feature and the upsampled recovery feature as a positive sample. It then compares the direction of the regular constraint fusion feature, thus reducing model complexity. Experimental results show that this Cross-Scale fusion method produces more accurate segmentation performance, improving the Dice index by 2%-3% on the ISIC2018 dataset and the SegPC2021 dataset." @default.
- W4387678029 created "2023-10-17" @default.
- W4387678029 creator A5011482849 @default.
- W4387678029 creator A5055204836 @default.
- W4387678029 creator A5080565326 @default.
- W4387678029 creator A5083081446 @default.
- W4387678029 date "2023-08-30" @default.
- W4387678029 modified "2023-10-17" @default.
- W4387678029 title "ECSFF: Exploring Efficient Cross-Scale Feature Fusion for Medical Image Segmentation" @default.
- W4387678029 cites W1901129140 @default.
- W4387678029 cites W1903029394 @default.
- W4387678029 cites W2138621090 @default.
- W4387678029 cites W3015788359 @default.
- W4387678029 cites W3035524453 @default.
- W4387678029 cites W3138516171 @default.
- W4387678029 cites W3173269149 @default.
- W4387678029 cites W3197957534 @default.
- W4387678029 cites W4386076603 @default.
- W4387678029 doi "https://doi.org/10.1109/icac57885.2023.10275282" @default.
- W4387678029 hasPublicationYear "2023" @default.
- W4387678029 type Work @default.
- W4387678029 citedByCount "0" @default.
- W4387678029 crossrefType "proceedings-article" @default.
- W4387678029 hasAuthorship W4387678029A5011482849 @default.
- W4387678029 hasAuthorship W4387678029A5055204836 @default.
- W4387678029 hasAuthorship W4387678029A5080565326 @default.
- W4387678029 hasAuthorship W4387678029A5083081446 @default.
- W4387678029 hasConcept C103038307 @default.
- W4387678029 hasConcept C110384440 @default.
- W4387678029 hasConcept C115961682 @default.
- W4387678029 hasConcept C119599485 @default.
- W4387678029 hasConcept C121332964 @default.
- W4387678029 hasConcept C127413603 @default.
- W4387678029 hasConcept C138885662 @default.
- W4387678029 hasConcept C141353440 @default.
- W4387678029 hasConcept C153180895 @default.
- W4387678029 hasConcept C154945302 @default.
- W4387678029 hasConcept C158525013 @default.
- W4387678029 hasConcept C173414695 @default.
- W4387678029 hasConcept C185592680 @default.
- W4387678029 hasConcept C198531522 @default.
- W4387678029 hasConcept C2776401178 @default.
- W4387678029 hasConcept C2778755073 @default.
- W4387678029 hasConcept C31972630 @default.
- W4387678029 hasConcept C41008148 @default.
- W4387678029 hasConcept C41895202 @default.
- W4387678029 hasConcept C43617362 @default.
- W4387678029 hasConcept C52622490 @default.
- W4387678029 hasConcept C62520636 @default.
- W4387678029 hasConcept C69744172 @default.
- W4387678029 hasConcept C89600930 @default.
- W4387678029 hasConceptScore W4387678029C103038307 @default.
- W4387678029 hasConceptScore W4387678029C110384440 @default.
- W4387678029 hasConceptScore W4387678029C115961682 @default.
- W4387678029 hasConceptScore W4387678029C119599485 @default.
- W4387678029 hasConceptScore W4387678029C121332964 @default.
- W4387678029 hasConceptScore W4387678029C127413603 @default.
- W4387678029 hasConceptScore W4387678029C138885662 @default.
- W4387678029 hasConceptScore W4387678029C141353440 @default.
- W4387678029 hasConceptScore W4387678029C153180895 @default.
- W4387678029 hasConceptScore W4387678029C154945302 @default.
- W4387678029 hasConceptScore W4387678029C158525013 @default.
- W4387678029 hasConceptScore W4387678029C173414695 @default.
- W4387678029 hasConceptScore W4387678029C185592680 @default.
- W4387678029 hasConceptScore W4387678029C198531522 @default.
- W4387678029 hasConceptScore W4387678029C2776401178 @default.
- W4387678029 hasConceptScore W4387678029C2778755073 @default.
- W4387678029 hasConceptScore W4387678029C31972630 @default.
- W4387678029 hasConceptScore W4387678029C41008148 @default.
- W4387678029 hasConceptScore W4387678029C41895202 @default.
- W4387678029 hasConceptScore W4387678029C43617362 @default.
- W4387678029 hasConceptScore W4387678029C52622490 @default.
- W4387678029 hasConceptScore W4387678029C62520636 @default.
- W4387678029 hasConceptScore W4387678029C69744172 @default.
- W4387678029 hasConceptScore W4387678029C89600930 @default.
- W4387678029 hasFunder F4320321001 @default.
- W4387678029 hasFunder F4320321540 @default.
- W4387678029 hasFunder F4320321921 @default.
- W4387678029 hasLocation W43876780291 @default.
- W4387678029 hasOpenAccess W4387678029 @default.
- W4387678029 hasPrimaryLocation W43876780291 @default.
- W4387678029 hasRelatedWork W1750358731 @default.
- W4387678029 hasRelatedWork W1971268144 @default.
- W4387678029 hasRelatedWork W2095903272 @default.
- W4387678029 hasRelatedWork W2139242969 @default.
- W4387678029 hasRelatedWork W2204403038 @default.
- W4387678029 hasRelatedWork W2391745328 @default.
- W4387678029 hasRelatedWork W2554790198 @default.
- W4387678029 hasRelatedWork W2788731446 @default.
- W4387678029 hasRelatedWork W2887520098 @default.
- W4387678029 hasRelatedWork W3152170969 @default.
- W4387678029 isParatext "false" @default.
- W4387678029 isRetracted "false" @default.
- W4387678029 workType "article" @default.