サッカー選手のアノテーションの品質向上を目指した深層学習のためのデータセットの構築およびパラメータの最適化に関する研究

• 資料種別: 記事
• タイトル: サッカー選手のアノテーションの品質向上を目指した深層学習のためのデータセットの構築およびパラメータの最適化に関する研究
• 著者・編者: 松尾 龍平; 田中 成典; 姜 文渊
• 著者標目: 松尾 龍平; 田中 成典; 姜 文渊
• 並列タイトル等: Application Research for Constructing Datasets and Optimizing Parameters with Deep Learning Aimed at Improving Quality of Annotation System on Soccer Players
• タイトル（掲載誌）: 写真測量とリモートセンシング = Journal of the Japan Society of Photogrammetry and Remote Sensing
• 巻号年月日等（掲載誌）: 64(3):2025
• 掲載巻: 64
• 掲載号: 3
• 掲載ページ: 74-93
• 掲載年月日（W3CDTF）: 2025
• ISSN（掲載誌）: 0285-5844
• ISSN-L（掲載誌）: 0285-5844
• 出版事項（掲載誌）: 東京 : 日本写真測量学会
• 出版地（国名コード）: JP
• 本文の言語コード: jpn
• NDLC: ZN1
• 対象利用者: 一般
• 所蔵機関: 国立国会図書館
• 請求記号: Z16-147
• 連携機関・データベース: 国立国会図書館 : 国立国会図書館雑誌記事索引; https://ndlsearch.ndl.go.jp
• 書誌ID（NDLBibID）: 034266951; http://id.ndl.go.jp/bib/034266951
• 整理区分コード: 632

• 要約等: <p>In recent years, object detection methods using deep learning have been utilized for player position analysis in sports information science, contributing to performance improvement. However, these methods require a lot of effort in creating training datasets. The annotation system developed by the authors previously improved the efficiency of datasets generation, but full automation was not achieved. Therefore, the present research proposes a semi-automatic generation method for training datasets using image processing technology and AI. And then, we implemented a method to refine the training dataset while building the unique detection model, taking into account the background difference and detection results from the AI model. Ultimately, it was confirmed that as the unique model is refined, it will also be able to semi-automatically generate datasets for learning, contributing to the improvement of the annotation system.</p>
• DOI: 10.4287/jsprs.64.74
• オンライン閲覧公開範囲: インターネット公開
• 連携機関・データベース: 科学技術振興機構 : J-STAGE; http://www.jstage.jst.go.jp

• 要約等: <p>In recent years, object detection methods using deep learning have been utilized for player position analysis in sports information science, contributing to performance improvement. However, these methods require a lot of effort in creating training datasets. The annotation system developed by the authors previously improved the efficiency of datasets generation, but full automation was not achieved. Therefore, the present research proposes a semi-automatic generation method for training datasets using image processing technology and AI. And then, we implemented a method to refine the training dataset while building the unique detection model, taking into account the background difference and detection results from the AI model. Ultimately, it was confirmed that as the unique model is refined, it will also be able to semi-automatically generate datasets for learning, contributing to the improvement of the annotation system.</p>
• DOI: 10.4287/jsprs.64.74; https://doi.org/10.4287/jsprs.64.74
• 関連情報（URI）: http://id.ndl.go.jp/bib/034266951; https://ndlsearch.ndl.go.jp/books/R000000004-I034266951; https://www.jstage.jst.go.jp/article/jsprs/64/3/64_74/_pdf
• 参照: YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors; https://cir.nii.ac.jp/crid/1360016978713277568; Lean Training Data Generation for Planar Object Detection Models in Unsteady Logistics Contexts; https://cir.nii.ac.jp/crid/1360026341338127104; Automatic Dataset Generation for Specific Object Detection; https://cir.nii.ac.jp/crid/1360026342289706880; Automatic Generation of Photorealistic Training Data for Detection of Industrial Components; https://cir.nii.ac.jp/crid/1360307815158885376; Robust player detection and tracking in broadcast soccer video based on enhanced particle filter; https://cir.nii.ac.jp/crid/1360307815480651136; LVIS: A Dataset for Large Vocabulary Instance Segmentation; https://cir.nii.ac.jp/crid/1360579817501593728; Few training data for Objection Detection; https://cir.nii.ac.jp/crid/1360589292127940096; Action Recognition in Basketball with Inertial Measurement Unit-Supported Vest; https://cir.nii.ac.jp/crid/1360870767882764672; Microsoft COCO: Common Objects in Context; https://cir.nii.ac.jp/crid/1361137044941077760; Improved adaptive Gaussian mixture model for background subtraction; https://cir.nii.ac.jp/crid/1361137045490768640; Quantifying positional and temporal movement patterns in professional rugby union using global positioning system; https://cir.nii.ac.jp/crid/1361981468449989632; Efficient adaptive density estimation per image pixel for the task of background subtraction; https://cir.nii.ac.jp/crid/1362825894549322496
• 連携機関・データベース: 国立情報学研究所 : CiNii Research; https://cir.nii.ac.jp/
• 提供元機関・データベース: Japan Link Center; https://japanlinkcenter.org/top; 雑誌記事索引データベース; https://ndlsearch.ndl.go.jp; Crossref; https://www.crossref.org
• 書誌ID（NDLBibID）: 034266951; http://id.ndl.go.jp/bib/034266951