{"metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00234180","sets":["1164:4619:11539:11642"]},"path":["11642"],"owner":"44499","recid":"234180","title":["人の注目領域を用いたProtoPFormerによる詳細画像識別の精度向上"],"pubdate":{"attribute_name":"公開日","attribute_value":"2024-05-08"},"_buckets":{"deposit":"1d84ad1a-3e6c-4393-964f-314bb4d7f945"},"_deposit":{"id":"234180","pid":{"type":"depid","value":"234180","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"人の注目領域を用いたProtoPFormerによる詳細画像識別の精度向上","author_link":["637620","637618","637619","637621"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"人の注目領域を用いたProtoPFormerによる詳細画像識別の精度向上"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"卒論スポットライトセッション (CVIM)","subitem_subject_scheme":"Other"}]},"item_type_id":"4","publish_date":"2024-05-08","item_4_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"中部大学"},{"subitem_text_value":"中部大学"},{"subitem_text_value":"中部大学"},{"subitem_text_value":"中部大学"}]},"item_4_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"Chubu University","subitem_text_language":"en"},{"subitem_text_value":"Chubu University","subitem_text_language":"en"},{"subitem_text_value":"Chubu University","subitem_text_language":"en"},{"subitem_text_value":"Chubu University","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/234180/files/IPSJ-CVIM24238049.pdf","label":"IPSJ-CVIM24238049.pdf"},"date":[{"dateType":"Available","dateValue":"2026-05-08"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-CVIM24238049.pdf","filesize":[{"value":"1.3 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":"20"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"f718195a-db91-4ac9-8881-513b52025373","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":"落合, 祐馬"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"平川, 翼"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"山下, 隆義"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"藤吉, 弘亘"}],"nameIdentifiers":[{}]}]},"item_4_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AA11131797","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":"2188-8701","subitem_source_identifier_type":"ISSN"}]},"item_4_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"プロトタイプベースのモデルはプロトタイプを用いて構築されるモデルであり，どのようなプロトタイプを用いて識別したかを示すことで，クラス特有の領域を可視化することができる．そのため，推論におけるモデルの解釈性が高い．プロトタイプは，各クラスに割り当てられる特徴的な領域を学習により獲得したものであり，入力画像と各クラスのプロトタイプの類似度から分類を行う．Proposes Prototypical Part Transformer (ProtoPFormer) [1] は Vision Transformer (ViT) [2] にプロトタイプを適用したモデルである．ViT による高い精度に加えて，各プロトタイプの注目領域が小さくなるように損失を与えるため，細かな領域に対する判断根拠を表示することができる．そこで，本研究では，詳細画像分類の精度向上を目的として ProtoPFormer に人の知見を導入する Branch を提案する．詳細画像分類ではクラス特有の領域に注目することが重要である．そこで，追加する Branch は人の知見に基づいてクラス特有の領域に注目するように設計された．本論文では，ProtoPFormer と本手法の精度を比較し，判断根拠を可視化したアテンションマップの注目領域を比較することで，本手法の有効性を評価する．","subitem_description_type":"Other"}]},"item_4_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"5","bibliographic_titles":[{"bibliographic_title":"研究報告コンピュータビジョンとイメージメディア（CVIM）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2024-05-08","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"49","bibliographicVolumeNumber":"2024-CVIM-238"}]},"relation_version_is_last":true,"weko_creator_id":"44499"},"id":234180,"updated":"2025-01-19T09:52:37.110515+00:00","links":{},"created":"2025-01-19T01:35:52.156361+00:00"}