{"id":206252,"updated":"2025-01-19T19:29:13.870528+00:00","links":{},"created":"2025-01-19T01:08:19.635528+00:00","metadata":{"_oai":{"id":"oai:ipsj.ixsq.nii.ac.jp:00206252","sets":["934:1022:9995:10238"]},"path":["10238"],"owner":"44499","recid":"206252","title":["敵対的学習に基づくドメイン適応によるドライブレコーダを用いたヒヤリハットの検出および分類"],"pubdate":{"attribute_name":"公開日","attribute_value":"2020-07-16"},"_buckets":{"deposit":"e5ce6a59-887e-4e44-bbd9-cac761683f98"},"_deposit":{"id":"206252","pid":{"type":"depid","value":"206252","revision_id":0},"owners":[44499],"status":"published","created_by":44499},"item_title":"敵対的学習に基づくドメイン適応によるドライブレコーダを用いたヒヤリハットの検出および分類","author_link":["512376","512373","512377","512375","512379","512381","512380","512378","512374","512372"],"item_titles":{"attribute_name":"タイトル","attribute_value_mlt":[{"subitem_title":"敵対的学習に基づくドメイン適応によるドライブレコーダを用いたヒヤリハットの検出および分類"},{"subitem_title":"Domain Adaptation based on Adversarial Learning for Detection and Classification of Near-miss Incident from Dashcam Data","subitem_title_language":"en"}]},"item_keyword":{"attribute_name":"キーワード","attribute_value_mlt":[{"subitem_subject":"[研究論文] 教師なしドメイン適用，ヒヤリハット，CNN，RNN","subitem_subject_scheme":"Other"}]},"item_type_id":"3","publish_date":"2020-07-16","item_3_text_3":{"attribute_name":"著者所属","attribute_value_mlt":[{"subitem_text_value":"日本電信電話株式会社NTTサービスエボリューション研究所"},{"subitem_text_value":"日本電信電話株式会社NTTサービスエボリューション研究所"},{"subitem_text_value":"日本電信電話株式会社NTTサービスエボリューション研究所"},{"subitem_text_value":"日本電信電話株式会社NTTサービスエボリューション研究所"},{"subitem_text_value":"日本電信電話株式会社NTTサービスエボリューション研究所"}]},"item_3_text_4":{"attribute_name":"著者所属(英)","attribute_value_mlt":[{"subitem_text_value":"NTT Service Evolution Laboratories, NTT Corporation","subitem_text_language":"en"},{"subitem_text_value":"NTT Service Evolution Laboratories, NTT Corporation","subitem_text_language":"en"},{"subitem_text_value":"NTT Service Evolution Laboratories, NTT Corporation","subitem_text_language":"en"},{"subitem_text_value":"NTT Service Evolution Laboratories, NTT Corporation","subitem_text_language":"en"},{"subitem_text_value":"NTT Service Evolution Laboratories, NTT 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/206252/files/IPSJ-TOD1303005.pdf","label":"IPSJ-TOD1303005.pdf"},"date":[{"dateType":"Available","dateValue":"2022-07-16"}],"format":"application/pdf","billing":["billing_file"],"filename":"IPSJ-TOD1303005.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":"13"},{"tax":["include_tax"],"price":"0","billingrole":"39"},{"tax":["include_tax"],"price":"0","billingrole":"44"}],"accessrole":"open_date","version_id":"489fdc0c-196c-4349-8744-e61d489e8f3a","displaytype":"detail","licensetype":"license_note","license_note":"Copyright (c) 2020 by the Information Processing Society of Japan"}]},"item_3_creator_5":{"attribute_name":"著者名","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"瀧本, 祥章"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"山本, 修平"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"松林, 達史"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"倉島, 健"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"戸田, 浩之"}],"nameIdentifiers":[{}]}]},"item_3_creator_6":{"attribute_name":"著者名(英)","attribute_type":"creator","attribute_value_mlt":[{"creatorNames":[{"creatorName":"Yoshiaki, Takimoto","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Shuhei, Yamamoto","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Tatsushi, Matsubayashi","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Takeshi, Kurashima","creatorNameLang":"en"}],"nameIdentifiers":[{}]},{"creatorNames":[{"creatorName":"Hiroyuki, Toda","creatorNameLang":"en"}],"nameIdentifiers":[{}]}]},"item_3_source_id_9":{"attribute_name":"書誌レコードID","attribute_value_mlt":[{"subitem_source_identifier":"AA11464847","subitem_source_identifier_type":"NCID"}]},"item_resource_type":{"attribute_name":"資源タイプ","attribute_value_mlt":[{"resourceuri":"http://purl.org/coar/resource_type/c_6501","resourcetype":"journal article"}]},"item_3_source_id_11":{"attribute_name":"ISSN","attribute_value_mlt":[{"subitem_source_identifier":"1882-7799","subitem_source_identifier_type":"ISSN"}]},"item_3_description_7":{"attribute_name":"論文抄録","attribute_value_mlt":[{"subitem_description":"ドライブレコーダによって収集された膨大なデータから交通事故やヒヤリハットなどの危険な状態を表すものを抽出，分類し，それらのデータを利活用する取り組みが近年行われている．しかしながら，それらのデータを自動的に検出，分類するには多量の教師データを人手で用意する必要があり，高コストである．また，既存の教師ありデータセットを用いて学習したモデルの活用も考えられる．しかし，車種やカメラの種類など，収集時の条件が異なると，データ集合全体の傾向が大きく変化するため，そのモデルによる検出や分類精度は制限される．そこで本稿では，分類対象のデータとそれとは異なる環境で収集された教師データがあるデータセットをもとに学習を行い，高精度な分類を行えるモデルの構築を目指す．具体的には，既存のConvolutional Recurrent Neural Networkベースのヒヤリハット分類手法をベースに，収集した環境を推定する層を追加することで，敵対的学習による教師なしドメイン適応を行う．これによって，従来は収集した環境に大きく依存した特徴量を抽出するConvolutional Neural Network部分について，環境に依存しない特徴量の抽出が期待できる．実験では，実際のドライブレコーダデータを用いて検証を行い，教師なしドメイン適応によって教師データがない環境下のデータ集合に対しても，高い分類精度を持つモデルを構築できることを示した．","subitem_description_type":"Other"}]},"item_3_description_8":{"attribute_name":"論文抄録(英)","attribute_value_mlt":[{"subitem_description":"Large amounts of highly useful driving data are being collected by dashboard cameras. By extracting dangerous situations from them and then classifying them into traffic accidents and near-miss incidents, many attractive applications are made possible. However, manually extraction is expensive because it requires a large amount of labeling. One method uses a model trained on an existing supervised dataset. If the collection conditions such as the model of vehicle and camera are different, the tendency of the dataset changes and the accuracy of the model is poor. In this paper, we propose a method that classifies target data with high accuracy by training based on two kinds of datasets: unlabeled target dataset, and labeled source dataset collected in an environment different from the target data. Specifically, our solution adds a layer to the existing convolutional recurrent neural network-based near-miss detection method. The added layer realizes unsupervised domain adaptation based on adversarial learning. As a result, the features extracted by the convolutional neural network are independent of the environment. An experiment on actual driving data verifies that the proposed method is effective.","subitem_description_type":"Other"}]},"item_3_biblio_info_10":{"attribute_name":"書誌情報","attribute_value_mlt":[{"bibliographicPageEnd":"9","bibliographic_titles":[{"bibliographic_title":"情報処理学会論文誌データベース（TOD）"}],"bibliographicPageStart":"1","bibliographicIssueDates":{"bibliographicIssueDate":"2020-07-16","bibliographicIssueDateType":"Issued"},"bibliographicIssueNumber":"3","bibliographicVolumeNumber":"13"}]},"relation_version_is_last":true,"weko_creator_id":"44499"}}