Matches in SemOpenAlex for { <https://semopenalex.org/work/W4386075569> ?p ?o ?g. }
Showing items 1 to 89 of
89
with 100 items per page.
- W4386075569 abstract "Model quantization is a crucial step for deploying super resolution (SR) networks on mobile devices. However, existing works focus on quantization-aware training, which requires complete dataset and expensive computational overhead. In this paper, we study post-training quantization (PTQ) for image super resolution using only a few unlabeled calibration images. As the SR model aims to maintain the texture and color information of input images, the distribution of activations are long-tailed, asymmetric and highly dynamic compared with classification models. To this end, we introduce the density-based dual clipping to cut off the outliers based on analyzing the asymmetric bounds of activations. Moreover, we present a novel pixel aware calibration method with the supervision of the full-precision model to accommodate the highly dynamic range of different samples. Extensive experiments demonstrate that the proposed method significantly outperforms existing PTQ algorithms on various models and datasets. For instance, we get a 2.091 dB increase on Urban100 benchmark when quantizing <tex xmlns:mml=http://www.w3.org/1998/Math/MathML xmlns:xlink=http://www.w3.org/1999/xlink>$EDSRtimes4$</tex> to 4-bit with 100 unlabeled images. Our code is available at both PyTorch and MindSpore." @default.
- W4386075569 created "2023-08-23" @default.
- W4386075569 creator A5025782908 @default.
- W4386075569 creator A5035602823 @default.
- W4386075569 creator A5062581018 @default.
- W4386075569 creator A5081216284 @default.
- W4386075569 date "2023-06-01" @default.
- W4386075569 modified "2023-10-16" @default.
- W4386075569 title "Toward Accurate Post-Training Quantization for Image Super Resolution" @default.
- W4386075569 cites W1930824406 @default.
- W4386075569 cites W1985806826 @default.
- W4386075569 cites W2016482162 @default.
- W4386075569 cites W2121927366 @default.
- W4386075569 cites W2133665775 @default.
- W4386075569 cites W2739757502 @default.
- W4386075569 cites W2895432151 @default.
- W4386075569 cites W2949080081 @default.
- W4386075569 cites W2956093359 @default.
- W4386075569 cites W2962298324 @default.
- W4386075569 cites W2963122961 @default.
- W4386075569 cites W2963372104 @default.
- W4386075569 cites W2963470893 @default.
- W4386075569 cites W2963606888 @default.
- W4386075569 cites W2970608956 @default.
- W4386075569 cites W2971595222 @default.
- W4386075569 cites W3015342281 @default.
- W4386075569 cites W3096641820 @default.
- W4386075569 cites W3110378228 @default.
- W4386075569 cites W3173515979 @default.
- W4386075569 cites W3177491619 @default.
- W4386075569 cites W3193148745 @default.
- W4386075569 cites W3203827541 @default.
- W4386075569 cites W3204989974 @default.
- W4386075569 cites W4213068294 @default.
- W4386075569 cites W4292787069 @default.
- W4386075569 cites W4312383953 @default.
- W4386075569 cites W4312400283 @default.
- W4386075569 cites W4312718416 @default.
- W4386075569 doi "https://doi.org/10.1109/cvpr52729.2023.00567" @default.
- W4386075569 hasPublicationYear "2023" @default.
- W4386075569 type Work @default.
- W4386075569 citedByCount "0" @default.
- W4386075569 crossrefType "proceedings-article" @default.
- W4386075569 hasAuthorship W4386075569A5025782908 @default.
- W4386075569 hasAuthorship W4386075569A5035602823 @default.
- W4386075569 hasAuthorship W4386075569A5062581018 @default.
- W4386075569 hasAuthorship W4386075569A5081216284 @default.
- W4386075569 hasConcept C13280743 @default.
- W4386075569 hasConcept C138885662 @default.
- W4386075569 hasConcept C153180895 @default.
- W4386075569 hasConcept C154945302 @default.
- W4386075569 hasConcept C160633673 @default.
- W4386075569 hasConcept C185798385 @default.
- W4386075569 hasConcept C205649164 @default.
- W4386075569 hasConcept C2776848632 @default.
- W4386075569 hasConcept C28855332 @default.
- W4386075569 hasConcept C31972630 @default.
- W4386075569 hasConcept C41008148 @default.
- W4386075569 hasConcept C41895202 @default.
- W4386075569 hasConcept C79337645 @default.
- W4386075569 hasConceptScore W4386075569C13280743 @default.
- W4386075569 hasConceptScore W4386075569C138885662 @default.
- W4386075569 hasConceptScore W4386075569C153180895 @default.
- W4386075569 hasConceptScore W4386075569C154945302 @default.
- W4386075569 hasConceptScore W4386075569C160633673 @default.
- W4386075569 hasConceptScore W4386075569C185798385 @default.
- W4386075569 hasConceptScore W4386075569C205649164 @default.
- W4386075569 hasConceptScore W4386075569C2776848632 @default.
- W4386075569 hasConceptScore W4386075569C28855332 @default.
- W4386075569 hasConceptScore W4386075569C31972630 @default.
- W4386075569 hasConceptScore W4386075569C41008148 @default.
- W4386075569 hasConceptScore W4386075569C41895202 @default.
- W4386075569 hasConceptScore W4386075569C79337645 @default.
- W4386075569 hasLocation W43860755691 @default.
- W4386075569 hasOpenAccess W4386075569 @default.
- W4386075569 hasPrimaryLocation W43860755691 @default.
- W4386075569 hasRelatedWork W121273120 @default.
- W4386075569 hasRelatedWork W2002009170 @default.
- W4386075569 hasRelatedWork W2034462085 @default.
- W4386075569 hasRelatedWork W2136485282 @default.
- W4386075569 hasRelatedWork W2141888456 @default.
- W4386075569 hasRelatedWork W2337415362 @default.
- W4386075569 hasRelatedWork W2546871836 @default.
- W4386075569 hasRelatedWork W2740820121 @default.
- W4386075569 hasRelatedWork W317572212 @default.
- W4386075569 hasRelatedWork W4312857205 @default.
- W4386075569 isParatext "false" @default.
- W4386075569 isRetracted "false" @default.
- W4386075569 workType "article" @default.