Matches in SemOpenAlex for { <https://semopenalex.org/work/W4386066710> ?p ?o ?g. }
- W4386066710 abstract "Recently, deep neural networks have been successfully applied for image restoration (IR) (e.g., super-resolution, de-noising, de-blurring). Despite their promising performance, running IR networks requires heavy computation. A large body of work has been devoted to addressing this issue by designing novel neural networks or pruning their parameters. However, the common limitation is that while images are saved in a compressed format before being enhanced by IR, prior work does not consider the impact of compression on the IR quality. In this paper, we present AccelIR, a framework that optimizes image compression considering the end-to-end pipeline of IR tasks. AccelIR encodes an image through IR-aware compression that optimizes compression levels across image blocks within an image according to the impact on the IR quality. Then, it runs a lightweight IR network on the compressed image, effectively reducing IR computation, while maintaining the same IR quality and image size. Our extensive evaluation using nine IR networks shows that AccelIR can reduce the computing overhead of super-resolution, de-nosing, and de-blurring by 49%, 29%, and 32% on average, respectively." @default.
- W4386066710 created "2023-08-23" @default.
- W4386066710 creator A5015247369 @default.
- W4386066710 creator A5019322974 @default.
- W4386066710 creator A5028488603 @default.
- W4386066710 creator A5029128283 @default.
- W4386066710 date "2023-06-01" @default.
- W4386066710 modified "2023-09-27" @default.
- W4386066710 title "AccelIR: Task-aware Image Compression for Accelerating Neural Restoration" @default.
- W4386066710 cites W1969483458 @default.
- W4386066710 cites W2099563019 @default.
- W4386066710 cites W2100095375 @default.
- W4386066710 cites W2122086266 @default.
- W4386066710 cites W2123113293 @default.
- W4386066710 cites W2140196014 @default.
- W4386066710 cites W2476548250 @default.
- W4386066710 cites W2503339013 @default.
- W4386066710 cites W2607041014 @default.
- W4386066710 cites W2613155248 @default.
- W4386066710 cites W2620296437 @default.
- W4386066710 cites W2741137940 @default.
- W4386066710 cites W2767850900 @default.
- W4386066710 cites W2792447253 @default.
- W4386066710 cites W2809226111 @default.
- W4386066710 cites W2891639355 @default.
- W4386066710 cites W2916004337 @default.
- W4386066710 cites W2949208911 @default.
- W4386066710 cites W2956093359 @default.
- W4386066710 cites W2963149687 @default.
- W4386066710 cites W2963372104 @default.
- W4386066710 cites W2963494934 @default.
- W4386066710 cites W2997572967 @default.
- W4386066710 cites W2999276659 @default.
- W4386066710 cites W3012118477 @default.
- W4386066710 cites W3034352949 @default.
- W4386066710 cites W3034469748 @default.
- W4386066710 cites W3035638508 @default.
- W4386066710 cites W3047354607 @default.
- W4386066710 cites W3098848838 @default.
- W4386066710 cites W3102451820 @default.
- W4386066710 cites W3104725225 @default.
- W4386066710 cites W3105328221 @default.
- W4386066710 cites W3167297682 @default.
- W4386066710 cites W3175317007 @default.
- W4386066710 cites W3176997885 @default.
- W4386066710 cites W3178925107 @default.
- W4386066710 cites W3193508667 @default.
- W4386066710 cites W3202040256 @default.
- W4386066710 cites W4239851455 @default.
- W4386066710 cites W4242059867 @default.
- W4386066710 cites W4312348045 @default.
- W4386066710 doi "https://doi.org/10.1109/cvpr52729.2023.01747" @default.
- W4386066710 hasPublicationYear "2023" @default.
- W4386066710 type Work @default.
- W4386066710 citedByCount "0" @default.
- W4386066710 crossrefType "proceedings-article" @default.
- W4386066710 hasAuthorship W4386066710A5015247369 @default.
- W4386066710 hasAuthorship W4386066710A5019322974 @default.
- W4386066710 hasAuthorship W4386066710A5028488603 @default.
- W4386066710 hasAuthorship W4386066710A5029128283 @default.
- W4386066710 hasConcept C106430172 @default.
- W4386066710 hasConcept C108010975 @default.
- W4386066710 hasConcept C111919701 @default.
- W4386066710 hasConcept C11413529 @default.
- W4386066710 hasConcept C115961682 @default.
- W4386066710 hasConcept C13481523 @default.
- W4386066710 hasConcept C154945302 @default.
- W4386066710 hasConcept C159985019 @default.
- W4386066710 hasConcept C180016635 @default.
- W4386066710 hasConcept C192562407 @default.
- W4386066710 hasConcept C199360897 @default.
- W4386066710 hasConcept C2779960059 @default.
- W4386066710 hasConcept C31972630 @default.
- W4386066710 hasConcept C41008148 @default.
- W4386066710 hasConcept C43521106 @default.
- W4386066710 hasConcept C45374587 @default.
- W4386066710 hasConcept C50644808 @default.
- W4386066710 hasConcept C55020928 @default.
- W4386066710 hasConcept C6557445 @default.
- W4386066710 hasConcept C86803240 @default.
- W4386066710 hasConcept C9417928 @default.
- W4386066710 hasConceptScore W4386066710C106430172 @default.
- W4386066710 hasConceptScore W4386066710C108010975 @default.
- W4386066710 hasConceptScore W4386066710C111919701 @default.
- W4386066710 hasConceptScore W4386066710C11413529 @default.
- W4386066710 hasConceptScore W4386066710C115961682 @default.
- W4386066710 hasConceptScore W4386066710C13481523 @default.
- W4386066710 hasConceptScore W4386066710C154945302 @default.
- W4386066710 hasConceptScore W4386066710C159985019 @default.
- W4386066710 hasConceptScore W4386066710C180016635 @default.
- W4386066710 hasConceptScore W4386066710C192562407 @default.
- W4386066710 hasConceptScore W4386066710C199360897 @default.
- W4386066710 hasConceptScore W4386066710C2779960059 @default.
- W4386066710 hasConceptScore W4386066710C31972630 @default.
- W4386066710 hasConceptScore W4386066710C41008148 @default.
- W4386066710 hasConceptScore W4386066710C43521106 @default.
- W4386066710 hasConceptScore W4386066710C45374587 @default.
- W4386066710 hasConceptScore W4386066710C50644808 @default.
- W4386066710 hasConceptScore W4386066710C55020928 @default.
- W4386066710 hasConceptScore W4386066710C6557445 @default.