Accelerating Deep Learning Tasks with Optimized GPU-assisted Image Decoding
In computer vision deep learning (DL) tasks, most of the input image datasets are stored in the JPEG format. These JPEG datasets need to be decoded before DL tasks are performed on them. We observe two problems in the current JPEG decoding procedures for DL tasks: (1) the decoding of image entropy d...
Gespeichert in:
| Veröffentlicht in: | Proceedings - International Conference on Parallel and Distributed Systems S. 274 - 281 |
|---|---|
| Hauptverfasser: | , , |
| Format: | Tagungsbericht |
| Sprache: | Englisch |
| Veröffentlicht: |
IEEE
01.12.2020
|
| Schlagworte: | |
| ISSN: | 2690-5965 |
| Online-Zugang: | Volltext |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Abstract | In computer vision deep learning (DL) tasks, most of the input image datasets are stored in the JPEG format. These JPEG datasets need to be decoded before DL tasks are performed on them. We observe two problems in the current JPEG decoding procedures for DL tasks: (1) the decoding of image entropy data in the decoder is performed sequentially, and this sequential decoding repeats with the DL iterations, which takes significant time; (2) Current parallel decoding methods under-utilize the massive hardware threads on GPUs. To reduce the image decoding time, we introduce a pre-scan mechanism to avoid the repeated image scanning in DL tasks. Our pre-scan generates boundary markers for entropy data so that the decoding can be performed in parallel. To cooperate with the existing dataset storage and caching systems, we propose two modes of the pre-scan mechanism: a compatible mode and a fast mode. The compatible mode does not change the image file structure so pre-scanned files can be stored back to disk for subsequent DL tasks. In comparison, the fast mode crafts a JPEG image into a binary format suitable for parallel decoding, which can be processed directly on the GPU. Since the GPU has thousands of hardware threads, we propose a fine-grained parallel decoding method on the pre-scanned dataset. The fine-grained parallelism utilizes the GPU effectively, and achieves speedups of around 1.5× over existing GPU-assisted image decoding libraries on real-world DL tasks. |
|---|---|
| AbstractList | In computer vision deep learning (DL) tasks, most of the input image datasets are stored in the JPEG format. These JPEG datasets need to be decoded before DL tasks are performed on them. We observe two problems in the current JPEG decoding procedures for DL tasks: (1) the decoding of image entropy data in the decoder is performed sequentially, and this sequential decoding repeats with the DL iterations, which takes significant time; (2) Current parallel decoding methods under-utilize the massive hardware threads on GPUs. To reduce the image decoding time, we introduce a pre-scan mechanism to avoid the repeated image scanning in DL tasks. Our pre-scan generates boundary markers for entropy data so that the decoding can be performed in parallel. To cooperate with the existing dataset storage and caching systems, we propose two modes of the pre-scan mechanism: a compatible mode and a fast mode. The compatible mode does not change the image file structure so pre-scanned files can be stored back to disk for subsequent DL tasks. In comparison, the fast mode crafts a JPEG image into a binary format suitable for parallel decoding, which can be processed directly on the GPU. Since the GPU has thousands of hardware threads, we propose a fine-grained parallel decoding method on the pre-scanned dataset. The fine-grained parallelism utilizes the GPU effectively, and achieves speedups of around 1.5× over existing GPU-assisted image decoding libraries on real-world DL tasks. |
| Author | Luo, Qiong Yan, Shengen Wang, Lipeng |
| Author_xml | – sequence: 1 givenname: Lipeng surname: Wang fullname: Wang, Lipeng email: lwangay@cse.ust.hk organization: HKUST,Department of Computer Science and Engineering,Hong Kong, China – sequence: 2 givenname: Qiong surname: Luo fullname: Luo, Qiong email: luo@cse.ust.hk organization: HKUST,Department of Computer Science and Engineering,Hong Kong, China – sequence: 3 givenname: Shengen surname: Yan fullname: Yan, Shengen email: yanshengen@sensetime.com organization: SenseTime Research,Shenzhen,China |
| BookMark | eNotjd1Kw0AUhFdRsKk-gSB5gcRz9ie7uQyprcFAC7bXZZOc1NU2DdmA6NMb0avhg_lmAnbVnTti7AEhRoT0scg32eJVIUiIOXCIAUCqCxag5gZT0DK5ZDOepBCpNFE3LPD-HaaiUDBjL1ld05EGO7ruEC6I-rAkO3S_tLX-w4efbnwL1_3oTu6bmnC12UXWe-fHCYqTPdBk1edmEm7ZdWuPnu7-c852y6dt_hyV61WRZ2XkptMxwlaQ5tI0CKbmRlpqsWqTtjIaeSUFkCSrNEINjTYJCa5a2aCoiGpeVVrM2f3friOifT-4kx2-9qlQKQcUP69rT00 |
| CODEN | IEEPAD |
| ContentType | Conference Proceeding |
| DBID | 6IE 6IL CBEJK RIE RIL |
| DOI | 10.1109/ICPADS51040.2020.00045 |
| DatabaseName | IEEE Electronic Library (IEL) Conference Proceedings IEEE Proceedings Order Plan All Online (POP All Online) 1998-present by volume IEEE Xplore All Conference Proceedings IEEE Electronic Library (IEL) IEEE Proceedings Order Plans (POP All) 1998-Present |
| DatabaseTitleList | |
| Database_xml | – sequence: 1 dbid: RIE name: IEEE Electronic Library (IEL) url: https://ieeexplore.ieee.org/ sourceTypes: Publisher |
| DeliveryMethod | fulltext_linktorsrc |
| Discipline | Computer Science |
| EISBN | 1728190746 9781728190747 |
| EISSN | 2690-5965 |
| EndPage | 281 |
| ExternalDocumentID | 9359201 |
| Genre | orig-research |
| GroupedDBID | 23M 29O 6IE 6IF 6IH 6IK 6IL 6IM 6IN AAJGR AAWTH ABLEC ADZIZ ALMA_UNASSIGNED_HOLDINGS BEFXN BFFAM BGNUA BKEBE BPEOZ CBEJK CHZPO IEGSK IPLJI OCL RIE RIL RNS |
| ID | FETCH-LOGICAL-i203t-1f3e7248d108c284aef1bf6fb8712b430e4ea5710c0d786e325f4d13beec2bb73 |
| IEDL.DBID | RIE |
| ISICitedReferencesCount | 3 |
| ISICitedReferencesURI | http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=000662964400033&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| IngestDate | Wed Aug 27 02:41:17 EDT 2025 |
| IsPeerReviewed | false |
| IsScholarly | true |
| Language | English |
| LinkModel | DirectLink |
| MergedId | FETCHMERGED-LOGICAL-i203t-1f3e7248d108c284aef1bf6fb8712b430e4ea5710c0d786e325f4d13beec2bb73 |
| PageCount | 8 |
| ParticipantIDs | ieee_primary_9359201 |
| PublicationCentury | 2000 |
| PublicationDate | 2020-Dec. |
| PublicationDateYYYYMMDD | 2020-12-01 |
| PublicationDate_xml | – month: 12 year: 2020 text: 2020-Dec. |
| PublicationDecade | 2020 |
| PublicationTitle | Proceedings - International Conference on Parallel and Distributed Systems |
| PublicationTitleAbbrev | ICPADS |
| PublicationYear | 2020 |
| Publisher | IEEE |
| Publisher_xml | – name: IEEE |
| SSID | ssj0020350 |
| Score | 2.1485918 |
| Snippet | In computer vision deep learning (DL) tasks, most of the input image datasets are stored in the JPEG format. These JPEG datasets need to be decoded before DL... |
| SourceID | ieee |
| SourceType | Publisher |
| StartPage | 274 |
| SubjectTerms | Decoding Deep learning GPU Graphics processing units Hardware heterogeneous processing image decoding Iterative decoding parallel processing Task analysis Transform coding |
| Title | Accelerating Deep Learning Tasks with Optimized GPU-assisted Image Decoding |
| URI | https://ieeexplore.ieee.org/document/9359201 |
| WOSCitedRecordID | wos000662964400033&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| hasFullText | 1 |
| inHoldings | 1 |
| isFullTextHit | |
| isPrint | |
| link | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LSwMxEA61ePBUtRXf5ODRtZvHvo6ltVqEumALvZUkOylFui19ePDXO9ldK4IXbyEQQmaSTL6ZfDOE3IkkNpyB8rjhypPaBp5iLotoLJlNmLAmLotNRMNhPJkkaY3c77kwAFB8PoMH1yxi-dnS7JyrrO1YpNyRtQ6iKCy5Wntw5SJkFQOY-Ul70E07vTfccNJHEMj9Iitn8KuESmFB-o3_zX1MWj9UPJrujcwJqUF-ShrftRhodTSb5KVjDFoQp898RnsAK1qlTp3Rkdq8b6jzuNJXvCEW80_I6FM69vDh7LSc0cECrxUcZZZulhYZ9x9H3WevKpTgzXHdW49ZARGXcYZSNmhvFFimbWg1oiGupfBBggrwLWH8LIpDEDywMmNCAxiudSTOSD1f5nBOqEKEE2CvCbmRmqvE4gIBhETkobUILkjTyWa6KnNhTCuxXP7dfUWOnPDL7x_XpL5d7-CGHJqP7Xyzvi0U-AWjFpye |
| linkProvider | IEEE |
| linkToHtml | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwlV1LTwIxEG4ImugJFYxve_Doyvax7O6RoAgBcRMh4Uba7pQQwyM8PPjrnS4rxsSLt6ZJ03Sm7fRr-81HyJ2II8MZKI8brjypbeAp5rKIRpLZmAlroq3YRNjrRcNhnBTI_Y4LAwDZ5zN4cMXsLT-dm427Kqs6Fil3ZK09p5yVs7V28Mq9keUcYObH1XYjqT--4ZSTPsJA7md5OYNfIipZDGmW_tf7Ean8kPFosgszx6QAsxNS-lZjoPniLJNO3RiMIc6jszF9BFjQPHnqmPbV6n1F3Z0rfcU9Yjr5hJQ-JwMPj87OzyltT3FjwVZm7nqpkEHzqd9oeblUgjfBca89ZgWEXEYp2tlgxFFgmbY1qxEPcS2FDxJUgKcJ46dhVAPBAytTJjSA4VqH4pQUZ_MZnBGqEOMEWGtq3EjNVWxxgABCIvbQWgTnpOxsM1pss2GMcrNc_F19Sw5a_ZfuqNvudS7JoXPE9jPIFSmulxu4JvvmYz1ZLW8yZ34B-7qf5w |
| openUrl | ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=proceeding&rft.title=Proceedings+-+International+Conference+on+Parallel+and+Distributed+Systems&rft.atitle=Accelerating+Deep+Learning+Tasks+with+Optimized+GPU-assisted+Image+Decoding&rft.au=Wang%2C+Lipeng&rft.au=Luo%2C+Qiong&rft.au=Yan%2C+Shengen&rft.date=2020-12-01&rft.pub=IEEE&rft.eissn=2690-5965&rft.spage=274&rft.epage=281&rft_id=info:doi/10.1109%2FICPADS51040.2020.00045&rft.externalDocID=9359201 |