Matches in SemOpenAlex for { <https://semopenalex.org/work/W4280490190> ?p ?o ?g. }
Showing items 1 to 93 of
93
with 100 items per page.
- W4280490190 abstract "This paper explores training efficient VGG-style super-resolution (SR) networks with the structural re-parameterization technique. The general pipeline of re-parameterization is to train networks with multi-branch topology first, and then merge them into standard 3x3 convolutions for efficient inference. In this work, we revisit those primary designs and investigate essential components for re-parameterizing SR networks. First of all, we find that batch normalization (BN) is important to bring training non-linearity and improve the final performance. However, BN is typically ignored in SR, as it usually degrades the performance and introduces unpleasant artifacts. We carefully analyze the cause of BN issue and then propose a straightforward yet effective solution. In particular, we first train SR networks with mini-batch statistics as usual, and then switch to using population statistics at the later training period. While we have successfully re-introduced BN into SR, we further design a new re-parameterizable block tailored for SR, namely RepSR. It consists of a clean residual path and two expand-and-squeeze convolution paths with the modified BN. Extensive experiments demonstrate that our simple RepSR is capable of achieving superior performance to previous SR re-parameterization methods among different model sizes. In addition, our RepSR can achieve a better trade-off between performance and actual running time (throughput) than previous SR methods. Codes are available at https://github.com/TencentARC/RepSR." @default.
- W4280490190 created "2022-05-22" @default.
- W4280490190 creator A5020601466 @default.
- W4280490190 creator A5023845728 @default.
- W4280490190 creator A5088475144 @default.
- W4280490190 date "2022-10-10" @default.
- W4280490190 modified "2023-09-30" @default.
- W4280490190 title "RepSR: Training Efficient VGG-style Super-Resolution Networks with Structural Re-Parameterization and Batch Normalization" @default.
- W4280490190 cites W1791560514 @default.
- W4280490190 cites W1930824406 @default.
- W4280490190 cites W2047920195 @default.
- W4280490190 cites W2163935418 @default.
- W4280490190 cites W2557227117 @default.
- W4280490190 cites W2739757502 @default.
- W4280490190 cites W2741137940 @default.
- W4280490190 cites W2747898905 @default.
- W4280490190 cites W2795024892 @default.
- W4280490190 cites W2866634454 @default.
- W4280490190 cites W2954930822 @default.
- W4280490190 cites W2963184176 @default.
- W4280490190 cites W2963372104 @default.
- W4280490190 cites W2963470893 @default.
- W4280490190 cites W2963729050 @default.
- W4280490190 cites W2963814095 @default.
- W4280490190 cites W2964101377 @default.
- W4280490190 cites W2964277374 @default.
- W4280490190 cites W2965669158 @default.
- W4280490190 cites W3011005573 @default.
- W4280490190 cites W3034247386 @default.
- W4280490190 cites W3105328221 @default.
- W4280490190 cites W3109684201 @default.
- W4280490190 cites W3168684807 @default.
- W4280490190 cites W3171038842 @default.
- W4280490190 cites W3206130237 @default.
- W4280490190 cites W3207918547 @default.
- W4280490190 doi "https://doi.org/10.1145/3503161.3547915" @default.
- W4280490190 hasPublicationYear "2022" @default.
- W4280490190 type Work @default.
- W4280490190 citedByCount "9" @default.
- W4280490190 countsByYear W42804901902023 @default.
- W4280490190 crossrefType "proceedings-article" @default.
- W4280490190 hasAuthorship W4280490190A5020601466 @default.
- W4280490190 hasAuthorship W4280490190A5023845728 @default.
- W4280490190 hasAuthorship W4280490190A5088475144 @default.
- W4280490190 hasBestOaLocation W42804901901 @default.
- W4280490190 hasConcept C113775141 @default.
- W4280490190 hasConcept C11413529 @default.
- W4280490190 hasConcept C136886441 @default.
- W4280490190 hasConcept C144024400 @default.
- W4280490190 hasConcept C149923435 @default.
- W4280490190 hasConcept C154945302 @default.
- W4280490190 hasConcept C155512373 @default.
- W4280490190 hasConcept C173608175 @default.
- W4280490190 hasConcept C19165224 @default.
- W4280490190 hasConcept C197129107 @default.
- W4280490190 hasConcept C199360897 @default.
- W4280490190 hasConcept C2776214188 @default.
- W4280490190 hasConcept C2777735758 @default.
- W4280490190 hasConcept C2908647359 @default.
- W4280490190 hasConcept C41008148 @default.
- W4280490190 hasConceptScore W4280490190C113775141 @default.
- W4280490190 hasConceptScore W4280490190C11413529 @default.
- W4280490190 hasConceptScore W4280490190C136886441 @default.
- W4280490190 hasConceptScore W4280490190C144024400 @default.
- W4280490190 hasConceptScore W4280490190C149923435 @default.
- W4280490190 hasConceptScore W4280490190C154945302 @default.
- W4280490190 hasConceptScore W4280490190C155512373 @default.
- W4280490190 hasConceptScore W4280490190C173608175 @default.
- W4280490190 hasConceptScore W4280490190C19165224 @default.
- W4280490190 hasConceptScore W4280490190C197129107 @default.
- W4280490190 hasConceptScore W4280490190C199360897 @default.
- W4280490190 hasConceptScore W4280490190C2776214188 @default.
- W4280490190 hasConceptScore W4280490190C2777735758 @default.
- W4280490190 hasConceptScore W4280490190C2908647359 @default.
- W4280490190 hasConceptScore W4280490190C41008148 @default.
- W4280490190 hasFunder F4320321001 @default.
- W4280490190 hasLocation W42804901901 @default.
- W4280490190 hasLocation W42804901902 @default.
- W4280490190 hasOpenAccess W4280490190 @default.
- W4280490190 hasPrimaryLocation W42804901901 @default.
- W4280490190 hasRelatedWork W2369673098 @default.
- W4280490190 hasRelatedWork W2386315983 @default.
- W4280490190 hasRelatedWork W2556154603 @default.
- W4280490190 hasRelatedWork W2787684247 @default.
- W4280490190 hasRelatedWork W3007689282 @default.
- W4280490190 hasRelatedWork W3035794324 @default.
- W4280490190 hasRelatedWork W3100971012 @default.
- W4280490190 hasRelatedWork W4287178710 @default.
- W4280490190 hasRelatedWork W4289148071 @default.
- W4280490190 hasRelatedWork W4297807321 @default.
- W4280490190 isParatext "false" @default.
- W4280490190 isRetracted "false" @default.
- W4280490190 workType "article" @default.