{"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00233506","sets":["1164:2822:11469:11529"]},"path":["11529"],"owner":"44499","recid":"233506","title":["Sliding Window実現方式の自動選択機構によるCNN推論処理の高効率化"],"pubdate":{"attribute_name":"公開日","attribute_value":"2024-03-14"},"_buckets":{"deposit":"b05341a9-103b-4c9b-b6e6-65344bf5d1e6"},"_deposit":{"id":"233506","pid":{"type":"depid","value":"233506","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"Sliding Window実現方式の自動選択機構によるCNN推論処理の高効率化","author_link":["634412","634410","634409","634408","634413","634411"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"Sliding Window実現方式の自動選択機構によるCNN推論処理の高効率化"},{"subitem_title":"Highly efficient CNN inference through automatic selection of sliding window implementation","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"回路・システム設計","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2024-03-14","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"日本電気株式会社"},{"subitem_text_value":"日本電気株式会社"},{"subitem_text_value":"日本電気株式会社"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"NEC Corporation","subitem_text_language":"en"},{"subitem_text_value":"NEC Corporation","subitem_text_language":"en"},{"subitem_text_value":"NEC Corporation","subitem_text_language":"en"}]},"item_language":{"attribute_name":"言語","attribute_value_mlt":[{"subitem_language":"jpn"}]},"item_publisher":{"attribute_name":"出版者","attribute_value_mlt":[{"subitem_publisher":"情報処理学会","subitem_publisher_language":"ja"}]},"publish_status":"0","weko_shared_id":-1,"item_file_price":{"attribute_name":"Billing file","attribute_type":"file","attribute_value_mlt":[{"url":{"url":"https://ipsj.ixsq.nii.ac.jp/record/233506/files/IPSJ-EMB24065029.pdf","label":"IPSJ-EMB24065029.pdf"},"date":[{"dateType":"Available","dateValue":"2026-03-14"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-EMB24065029.pdf","filesize":[{"value":"1.2 MB"}],"mimetype":"application/pdf","priceinfo":[{"tax":["include_tax"],"price":"660","billingrole":"5"},{"tax":["include_tax"],"price":"330","billingrole":"6"},{"tax":["include_tax"],"price":"0","billingrole":"42"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"635359f6-be9e-421b-886e-33a6510dfc0f","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2024 by the Information Processing Society of Japan"}]},"item_4_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"酒井, 完"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"中村, 寿彦"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"阪, 慎一"}],"nameIdentifiers":[{}]}]},"item_4_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Tamotsu, Sakai","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Toshihiko, Nakamura","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Shinich, Saka","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AA12149313","subitem_source_identifier_type":"NCID"}]},"item_4_textarea_12":{"attribute_name":"Notice","attribute_value_mlt":[{"subitem_textarea_value":"SIG Technical Reports are nonrefereed and hence may later appear in any journals, conferences, symposia, etc."}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourceuri":"http://purl.org/coar/resource_type/c_18gh","resourcetype":"technical report"}]},"item_4_source_id_11":{"attribute_name":"ISSN","attribute_value_mlt":[{"subitem_source_identifier":"2188-868X","subitem_source_identifier_type":"ISSN"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"近年，様々な画像認識タスクにおいて広く使用されている CNN は認識精度を更に向上させる為，モデル規模を複雑化・巨大化させる傾向にあるが，それに伴い，組込み機器上においては高効率な CNN 推論処理が求められている．本稿では，FPGA 上の高効率な CNN 推論処理の実現に向け，既に提案されている FPGA 上の CNN デザイン自動生成フローで生成される HW アーキテクチャの改善を検討した．具体的には，畳み込み層やプーリング層を構成する Sliding Window ユニットに改善余地があることに着目し，低スループット用 Sliding Window ユニットの検討を行った上で，各層の要求スループットに応じて適切な Sliding Window 実現方式を自動選択する機構を，本フローを構成す るC++ ベースの CNN ライブラリに新たに導入した．評価の結果，生成された CNN デザインのスループットが変化することなく面積効率が改善することを確認し，Sliding Window 実現方式の自動選択機構の有用性を実証した．","subitem_description_type":"Other"}]},"item_4_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"Recently, Convolutional Neural Networks (CNNs) have been widely used in various range of image recognition tasks. To further improve recognition accuracy, the complexity and size of CNN models are more and more increasing as a result of which the demand for highly efficient CNN inference is required in embedded devices. In this paper, to achieve highly efficient CNN inference on FPGA, we have explored the improvement of hardware architectures generated by the proposed FPGA-based CNN design automation flow. After presenting dedicated sliding window architecture which operates at low throughput, we have introduced a system to automatically select the appropriate sliding window architecture in the C++ based CNN library which constitute the proposed design flow. Through evaluation, we have confirmed that the area efficiency has been improved without affecting the throughput of the generated CNN design, which demonstrates the effectiveness of the automatic selection system for Sliding Window implementation.","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"7","bibliographic_titles":[{"bibliographic_title":"研究報告組込みシステム（EMB）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2024-03-14","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"29","bibliographicVolumeNumber":"2024-EMB-65"}]},"relation_version_is_last":true,"weko_creator_id":"44499"},"id":233506,"updated":"2025-01-19T10:04:26.340276+00:00","links":{},"created":"2025-01-19T01:34:59.496916+00:00"}