|
Affiliation Department |
Department of Computer Science
|
|
Title |
Professor |
|
Homepage |
|
|
External Link |
|
TAMAKI Toru
|
|
|
Research Areas
-
Informatics / Perceptual information processing / コンピュータビジョン
External Career
-
Niigata University Research Assistant
2001.04 - 2005.09
Country：Japan
-
Hiroshima University Associate Professor
2005.10 - 2020.10
Country：Japan
-
ESIEE Paris, France Laboratoire d'Informatique Gaspard-Monge (LIGM), Équipe Algorithmes, Architectures, Analyse et Synthèse d'images (A3SI) chercheur associé
2015.05 - 2016.01
Country：France
Papers
-
ObjectMix: Data Augmentation by Copy-Pasting Objects in Videos for Action Recognition Reviewed International journal
Jun Kimata, Tomoya Nitta, Toru Tamaki
ACM MM 2022 Asia (MMAsia '22) 2022.12
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (international conference proceedings)
-
Temporal Cross-attention for Action Recognition Reviewed International journal
Ryota Hashiguchi, Toru Tamaki
2022.12
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (international conference proceedings)
Feature shifts have been shown to be useful for action recognition with CNN-based models since Temporal Shift Module (TSM) was proposed. It is based on frame-wise feature extraction with late fusion, and layer features are shifted along the time direction for the temporal interaction. TokenShift, a recent model based on Vision Transformer (ViT), also uses the temporal feature shift mechanism, which, however, does not fully exploit the structure of Multi-head Self-Attention (MSA) in ViT. In this paper, we propose Multi-head Self/Cross-Attention (MSCA), which fully utilizes the attention structure. TokenShift is based on a frame-wise ViT with features temporally shifted with successive frames (at time t+1 and t-1). In contrast, the proposed MSCA replaces MSA in the frame-wise ViT, and some MSA heads attend to successive frames instead of the current frame. The computation cost is the same as the frame-wise ViT and TokenShift as it simply changes the target to which the attention is taken. There is a choice about which of key, query, and value are taken from the successive frames, then we experimentally compared these variants with Kinetics400. We also investigate other variants in which the proposed MSCA is used along the patch dimension of ViT, instead of the head dimension. Experimental results show that a variant, MSCA-KV, shows the best performance and is better than TokenShift by 0.1% and then ViT by 1.2%.
Other Link： https://openaccess.thecvf.com/menu_other.html
-
Model-agnostic Multi-Domain Learning with Domain-Specific Adapters for Action Recognition Reviewed International journal
Kazuki Omi, Jun Kimata, Toru Tamaki
IEICE Transactions on Information and Systems E105-D ( 12 ) 2022.12
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：IEICE
In this paper, we propose a multi-domain learning model for action recognition. The proposed method inserts domain-specific adapters between layers of domain-independent layers of a backbone net- work. Unlike a multi-head network that switches classification heads only, our model switches not only the heads, but also the adapters for facilitating to learn feature representations universal to multiple domains. Unlike prior works, the proposed method is model-agnostic and doesn’t assume model structures unlike prior works. Experimental results on three popular action recognition datasets (HMDB51, UCF101, and Kinetics-400) demonstrate that the proposed method is more effective than a multi-head architecture and more efficient than separately training models for each domain.
DOI： 10.1587/transinf.2022EDP7058
Other Link： https://search.ieice.org/bin/summary_advpub.php?id=2022EDP7058&category=D&lang=E&abst=
-
動作行動認識の最前線:手法，タスク，データセット Invited
玉木徹
画像応用技術専門委員会 研究会報告 34 ( 4 ) 1 - 20 2022.11
Authorship：Lead author, Last author, Corresponding author Language：Japanese Publishing type：Research paper (conference, symposium, etc.)
Other Link： http://www.tc-iaip.org/research/
-
Performance Evaluation of Action Recognition Models on Low Quality Videos Reviewed International journal
Aoi Otani, Ryota Hashiguchi, Kazuki Omi, Norishige Fukushima, Toru Tamaki
IEEE Access 10 94898 - 94907 2022.09
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (scientific journal) Publisher：IEEE
In the design of action recognition models, the quality of videos is an important issue; however, the trade-off between the quality and performance is often ignored. In general, action recognition models are trained on high-quality videos, hence it is not known how the model performance degrades when tested on low-quality videos, and how much the quality of training videos affects the performance. The issue of video quality is important, however, it has not been studied so far. The goal of this study is to show the trade-off between the performance and the quality of training and test videos by quantitative performance evaluation of several action recognition models for transcoded videos in different qualities. First, we show how the video quality affects the performance of pre-trained models. We transcode the original validation videos of Kinetics400 by changing quality control parameters of JPEG (compression strength) and H.264/AVC (CRF). Then we use the transcoded videos to validate the pre-trained models. Second, we show how the models perform when trained on transcoded videos. We transcode the original training videos of Kinetics400 by changing the quality parameters of JPEG and H.264/AVC. Then we train the models on the transcoded training videos and validate them with the original and transcoded validation videos. Experimental results with JPEG transcoding show that there is no severe performance degradation (up to −1.5%) for compression strength smaller than 70 where no quality degradation is visually observed, and for larger than 80 the performance degrades linearly with respect to the quality index. Experiments with H.264/AVC transcoding show that there is no significant performance loss (up to −1%) with CRF30 while the total size of video files is reduced to 30%. In summary, the video quality doesn’t have a large impact on the performance of action recognition models unless the quality degradation is severe and visible. This enables us to transcode the tr...
DOI： 10.1109/ACCESS.2022.3204755
Other Link： https://ieeexplore.ieee.org/document/9878331
-
Object-ABN: Learning to Generate Sharp Attention Maps for Action Recognition International journal
Tomoya Nitta, Tsubasa Hirakawa, Hironobu Fujiyoshi, Toru Tamaki
2022.07
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (other academic)
In this paper we propose an extension of the Attention Branch Network (ABN) by using instance segmentation for generating sharper attention maps for action recognition. Methods for visual explanation such as Grad-CAM usually generate blurry maps which are not intuitive for humans to understand, particularly in recognizing actions of people in videos. Our proposed method, Object-ABN, tackles this issue by introducing a new mask loss that makes the generated attention maps close to the instance segmentation result. Further the PC loss and multiple attention maps are introduced to enhance the sharpness of the maps and improve the performance of classification. Experimental results with UCF101 and SSv2 shows that the generated maps by the proposed method are much clearer qualitatively and quantitatively than those of the original ABN.
DOI： 10.48550/arXiv.2207.13306
Other Link： https://doi.org/10.48550/arXiv.2207.13306
-
On the Performance Evaluation of Action Recognition Models on Transcoded Low Quality Videos International journal
Aoi Otani, Ryota Hashiguchi, Kazuki Omi, Norishige Fukushima, Toru Tamaki
2022.04
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (other academic)
In the design of action recognition models, the quality of videos in the dataset is an important issue, however the trade-off between the quality and performance is often ignored. In general, action recognition models are trained and tested on high-quality videos, but in actual situations where action recognition models are deployed, sometimes it might not be assumed that the input videos are of high quality. In this study, we report qualitative evaluations of action recognition models for the quality degradation associated with transcoding by JPEG and H.264/AVC. Experimental results are shown for evaluating the performance of pre-trained models on the transcoded validation videos of Kinetics400. The models are also trained on the transcoded training videos. From these results, we quantitatively show the degree of degradation of the model performance with respect to the degradation of the video quality.
DOI： 10.48550/arXiv.2204.09166
Other Link： https://doi.org/10.48550/arXiv.2204.09166
-
Model-agnostic Multi-Domain Learning with Domain-Specific Adapters for Action Recognition International journal
Kazuki Omi, Toru Tamaki
2022.04
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (other academic)
In this paper, we propose a multi-domain learning model for action recognition. The proposed method inserts domain-specific adapters between layers of domain-independent layers of a backbone network. Unlike a multi-head network that switches classification heads only, our model switches not only the heads, but also the adapters for facilitating to learn feature representations universal to multiple domains. Unlike prior works, the proposed method is model-agnostic and doesn't assume model structures unlike prior works. Experimental results on three popular action recognition datasets (HMDB51, UCF101, and Kinetics-400) demonstrate that the proposed method is more effective than a multi-head architecture and more efficient than separately training models for each domain.
DOI： 10.48550/arXiv.2204.07270
Other Link： https://doi.org/10.48550/arXiv.2204.07270
-
Vision Transformer with Cross-attention by Temporal Shift for Efficient Action Recognition International journal
Ryota Hashiguchi, Toru Tamaki
2022.04
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (other academic)
We propose Multi-head Self/Cross-Attention (MSCA), which introduces a temporal cross-attention mechanism for action recognition, based on the structure of the Multi-head Self-Attention (MSA) mechanism of the Vision Transformer (ViT). Simply applying ViT to each frame of a video frame can capture frame features, but cannot model temporal features. However, simply modeling temporal information with CNN or Transfomer is computationally expensive. TSM that perform feature shifting assume a CNN and cannot take advantage of the ViT structure. The proposed model captures temporal information by shifting the Query, Key, and Value in the calculation of MSA of ViT. This is efficient without additional coinformationmputational effort and is a suitable structure for extending ViT over temporal. Experiments on Kineitcs400 show the effectiveness of the proposed method and its superiority over previous methods.
DOI： 10.48550/arXiv.2204.00452
Other Link： https://doi.org/10.48550/arXiv.2204.00452
-
ObjectMix: Data Augmentation by Copy-Pasting Objects in Videos for Action Recognition International journal
Jun Kimata, Tomoya Nitta, Toru Tamaki
2022.04
Authorship：Last author, Corresponding author Language：English Publishing type：Research paper (other academic)
In this paper, we propose a data augmentation method for action recognition using instance segmentation. Although many data augmentation methods have been proposed for image recognition, few methods have been proposed for action recognition. Our proposed method, ObjectMix, extracts each object region from two videos using instance segmentation and combines them to create new videos. Experiments on two action recognition datasets, UCF101 and HMDB51, demonstrate the effectiveness of the proposed method and show its superiority over VideoMix, a prior work.
DOI： 10.48550/arXiv.2204.00239
Other Link： https://doi.org/10.48550/arXiv.2204.00239
Books and Other Publications
-
玉木徹, 小出哲士, 吉田成人（ Role： Contributor , Chapter 12 下部拡大内視鏡（NBI）AI ②）
オーム社 2022.11 （ ISBN:978-4-274-22564-2 ）
Total pages：250 Responsible for pages：8 Language：jpn Book type：Scholarly book
-
Tetsushi Koide, Toru Tamaki, Shigeto Yoshida, and Shinji Tanaka（ Role： Joint author , Chapter: A Hierarchical Type Segmentation Hardware for Colorectal Endoscopic Images with Narrow Band Imaging Magnification）
Jenny Stanford Publishing 2021.11 （ ISBN:9789814877633 ）
Total pages：396 Responsible for pages：21 Language：eng Book type：Scholarly book
Several developed countries are facing serious problems in medical environments owing to the aging society, and extension of healthy lifetime has become a big challenge. Biomedical engineering, in addition to life sciences and medicine, can help tackle these problems. Innovative technologies concerning minimally invasive treatment, prognosis and early diagnosis, point-of-care testing, regenerative medicine, and personalized medicine need to be developed to realize a healthy aging society.
This book presents cutting-edge research in biomedical engineering from materials, devices, imaging, and information perspectives. The contributors are senior members of the Research Center for Biomedical Engineering, supported by the Ministry of Education, Culture, Sports, Science and Technology, Japan. All chapters are results of collaborative research in engineering and life sciences and cover nanotechnology, materials, optical sensing technology, imaging technology, image processing technology, and biomechanics, all of which are important areas in biomedical engineering. The book will be a useful resource for researchers, students, and readers who are interested in biomedical engineering.Other Link： https://www.amazon.co.jp/Biomedical-Engineering-Akihiro-Miyauchi/dp/9814877638/
-
玉木徹（ Role： Sole translator）
講談社 2021.03 （ ISBN:978-4-06-516196-8 ）
Total pages：536 Responsible for pages：536 Language：jpn Book type：Scholarly book
-
人工知能学会 編（ Role： Contributor , 9-3 パターン認識・理解（グラフィカルモデル））
共立出版 2017.07 （ ISBN:978-4-320-12420-2 ）
Total pages：1580 Responsible for pages：709-713 Language：jpn Book type：Scholarly book
-
Pythonで体験するベイズ推論 : PyMCによるMCMC入門
Davidson-Pilon Cameron, 玉木 徹（ Role： Sole translator）
森北出版 2017 （ ISBN:9784627077911 ）
Language：jpn Book type：Scholarly book
-
スパースモデリング : l1/l0ノルム最小化の基礎理論と画像処理への応用
Elad Michael, 玉木 徹（ Role： Sole translator）
共立出版 2016 （ ISBN:9784320123946 ）
Language：jpn Book type：Scholarly book
-
Szeliski Richard, 玉木 徹（ Role： Joint translator）
共立出版 2013 （ ISBN:9784320123281 ）
Language：jpn Book type：Scholarly book
Misc
-
効率的な動作認識のためのシフトによる時間的な相互アテンションを用いたVision Transformer Invited
橋口凌大, 玉木 徹
2023.02
Authorship：Last author, Corresponding author Language：Japanese Publishing type：Article, review, commentary, editorial, etc. (trade magazine, newspaper, online media)
-
移動軌跡のデータサイエンス Invited
玉木徹
74 ( 2 ) 236 - 240 2020.03
Language：Japanese Publishing type：Article, review, commentary, editorial, etc. (trade magazine, newspaper, online media) Publisher：株式会社エヌ・ティー・エス
Other Link： http://www.nts-book.co.jp/item/detail/summary/bio/20051225_42bk.html
Presentations
-
卓球映像中の選手の死角における打球コース推定手法の提案
加藤祥真, 鬼頭明, 玉木徹, 澤野弘明
映像情報メディア学会 2022年冬季大会 2022.12 映像情報メディア学会
Event date： 2022.12
Language：Japanese Presentation type：Oral presentation (general)
Venue：東京理科大学, 東京・オンライン Country：Japan
-
加藤祥真, 鬼頭明, 玉木徹, 澤野弘明
第20回情報学ワークショップ 2022.12 WiNF事務局
Event date： 2022.12
Language：Japanese Presentation type：Oral presentation (general)
Venue：愛知工業大学, 愛知 Country：Japan
-
ObjectMix: Data Augmentation by Copy-Pasting Objects in Videos for Action Recognition
Jun Kimata, Tomoya Nitta, Toru Tamaki
ACM MM 2022 Asia (MMAsia '22) 2022.12 ACM MM 2022 Asia
Event date： 2022.12
Language：English Presentation type：Oral presentation (general)
Country：Japan
-
加藤祥真, 鬼頭明, 玉木徹, 澤野弘明
映像情報メディア学会スポーツ情報処理研究会（SIP） 2022.12 映像情報メディア学会スポーツ情報処理研究会（SIP）
Event date： 2022.12
Language：Japanese Presentation type：Oral presentation (general)
Venue：名古屋工業大学, 愛知 Country：Japan
-
Temporal Cross-attention for Action Recognition International conference
Ryota Hashiguchi, Toru Tamaki
ACCV2022 Workshop on Vision Transformers: Theory and applications (VTTA-ACCV2022) 2022.12
Event date： 2022.12
Language：English
Venue：Galaxy Macau, Macau Country：Macao
Feature shifts have been shown to be useful for action recognition with CNN-based models since Temporal Shift Module (TSM) was proposed. It is based on frame-wise feature extraction with late fusion, and layer features are shifted along the time direction for the temporal interaction. TokenShift, a recent model based on Vision Transformer (ViT), also uses the temporal feature shift mechanism, which, however, does not fully exploit the structure of Multi-head Self-Attention (MSA) in ViT. In this paper, we propose Multi-head Self/Cross-Attention (MSCA), which fully utilizes the attention structure. TokenShift is based on a frame-wise ViT with features temporally shifted with successive frames (at time t+1 and t-1). In contrast, the proposed MSCA replaces MSA in the frame-wise ViT, and some MSA heads attend to successive frames instead of the current frame. The computation cost is the same as the frame-wise ViT and TokenShift as it simply changes the target to which the attention is taken. There is a choice about which of key, query, and value are taken from the successive frames, then we experimentally compared these variants with Kinetics400. We also investigate other variants in which the proposed MSCA is used along the patch dimension of ViT, instead of the head dimension. Experimental results show that a variant, MSCA-KV, shows the best performance and is better than TokenShift by 0.1% and then ViT by 1.2%.
Other Link： https://sites.google.com/view/vtta-accv2022
-
Action recognition with generated sequences
Taiki Sugiura, Toru Tamaki
The 7th International Symposium on Biomedical Engineering (ISBE2022) 2022.11
Event date： 2022.11
Language：English Presentation type：Oral presentation (general)
Country：Japan
-
動作認識の最前線：手法，タスク，データセット Invited
玉木徹
精密工学会 画像応用技術専門委員会(IAIP)2022年度第4回定例研究会 2022.11 精密工学会 画像応用技術専門委員会(IAIP)
Event date： 2022.11
Language：Japanese Presentation type：Oral presentation (invited, special)
Venue：中央大学, 東京・オンライン Country：Japan
人物の行動認識（action recognition）はコンピュータビジョンの重要なトピックの一つです。本講演では，代表的な手法やタスク，データセットなどを俯瞰し，最新の研究についても紹介します。
-
卓球競技映像におけるスイング動作区間推定手法の提案
加藤祥真, 鬼頭明, 玉木徹, 澤野弘明
令和四年度 電気・電子・情報関係学会 東海支部連合大会 2022.08
Event date： 2022.08
Language：Japanese Presentation type：Poster presentation
Venue：オンライン Country：Japan
-
ObjectMix：動画像中の物体のコピー・ペーストによる動作認識のためのデータ拡張
木全潤, 仁田智也, 玉木 徹
第28回画像センシングシンポジウム(SSII2022) 2022.06 画像センシング技術研究会
Event date： 2022.06
Language：Japanese Presentation type：Poster presentation
Venue：パシフィコ横浜, 神奈川 Country：Japan
-
深層学習を用いたNICE/JNET分類に基づく大腸内視鏡画像診断支援の一手法
片山大輔, 呉泳飛, 道田竜一, 小出哲士, 玉木徹, 吉田成人, 岡本由貴, 岡志郎, 田中信治
第28回画像センシングシンポジウム(SSII2022) 2022.06 画像センシング技術研究会
Event date： 2022.06
Language：Japanese Presentation type：Poster presentation
Venue：パシフィコ横浜, 神奈川 Country：Japan
Industrial Property Rights
-
細幸広, 藤原翔, 船原佑介, 玉木徹
Application no：特願2019-213340 Date applied：2019.11
Announcement no：特開2021-85178 Date announced：2021.06
Patent/Registration no：特許第7246294号 Date registered：2023.03 Date issued：2023.03
Rights holder：コベルコ建機株式会社, 国立大学法人広島大学 Country of applicant：Domestic Country of acquisition：Domestic
【発明の詳細な説明】
【技術分野】
【０００１】
本発明は、腕部材に対して回転可能に取り付けられた容器の収容物の体積を計測する技術に関するものである。
【背景技術】
【０００２】
油圧ショベルにおいては、作業当日の作業量を把握するために、バケットが掘削した掘削物の体積が計算される。また、油圧ショベルが掘削物をダンプカーに積み込む作業を行うに場合、掘削物の体積がダンプカーの上限積載量を超えないように掘削物の体積が計算される。このように、掘削物の体積は、種々の用途に適用可能であるため、高精度に計算されることが望ましい。掘削物の体積を計算する技術として、下記の特許文献１、２が知られている。
【０００３】
特許文献１には、掘削後のバケットの状況を撮影した画像から算出されたバケットの表面形状と、排土後のバケット内の状況を撮影した画像から算出したバケットの内部形状との差を演算することにより、バケットの作業量を算出する技術が開示されている。
【０００４】
特許文献２には、掘削物が入った状態でバケットの開口面から掘削物表面までの長さと、バケットが空の時のバケットの底からバケットの開口面までの長さとを足すことにより、バケットの底から掘削物の表面までの長さを求め、この長さに基づいて掘削物の体積を計算する技術が開示されている。 -
川村 健介, 玉木 徹, 小櫃 剛人, 黒川 勇三
Applicant：国立大学法人広島大学
Application no：特願2014-188656 Date applied：2014.09
Announcement no：特開2016-059300 Date announced：2016.04
Country of applicant：Domestic Country of acquisition：Domestic
-
小出 哲士, ホアン アイン トゥワン, 吉田 成人, 三島 翼, 重見 悟, 玉木 徹, 平川 翼, 宮木 理恵, 杉 幸樹
Applicant：国立大学法人広島大学
Application no：特願2014-022425 Date applied：2014.02
Announcement no：特開2015-146970 Date announced：2015.08
Country of applicant：Domestic Country of acquisition：Domestic
-
田中 慎也, 土谷 千加夫, 玉木 徹, 栗田 多喜夫
Applicant：日産自動車株式会社, 国立大学法人広島大学
Application no：特願2012-267267 Date applied：2012.12
Announcement no：特開2014-115706 Date announced：2014.06
Country of applicant：Domestic Country of acquisition：Domestic
-
玉木 徹, 山村 毅, 大西 昇
Applicant：理化学研究所
Application no：特願2001-054686 Date applied：2001.02
Announcement no：特開2002-158915 Date announced：2002.05
Patent/Registration no：特許第3429280号 Date registered：2003.05 Date issued：2003.05
Country of applicant：Domestic Country of acquisition：Domestic
Awards
-
2020年度IPSJ-CGVI優秀研究発表賞
2021.06 情報処理学会コンピュータグラフィックスとビジュアル情報学研究発表会 スペクトル類似度を考慮した深層学習によるRGB画像からスペクトル画像への変換手法
坂本真啓, 金田和文, 玉木徹, Bisser Raytchev
Award type：International academic award (Japan or overseas) Country：Japan
-
電子情報通信学会情報・システムソサイエティ功労賞
2021.06 電子情報通信学会情報・システムソサイエティ
玉木徹
Award type：Award from Japanese society, conference, symposium, etc. Country：Japan
-
平成17年度金森奨励賞
2006.06 医用画像情報学会
玉木徹
Award type：Honored in official journal of a scientific society, scientific journal
-
平成11年度学生奨励賞
1999.11 電子情報通信学会東海支部
玉木徹
Scientific Research Funds Acquisition Results
-
Grant number：22K12090 2022.04 - 2025.03
科学研究費補助金 基盤研究(C)
玉木徹
Authorship：Principal investigator Grant type：Competitive
Grant amount：\4160000 （ Direct Cost: \3200000 、 Indirect Cost：\960000 ）
本研究の目的は，動画像理解のための時空間特徴量を取得する新しい方法論を構築することである．様々な動画像認識において空間的な情報と時間的な情報を，時空間情報としてひとまとめで扱う事が多いが，本研究が目指すのは，空間情報と時間情報を高いレベルで分離するというアプローチである．単に別々に特徴量を抽出するのではなく，様々な動画認識タスクに応用するために，時間と空間の情報を関連させつつ分離するために，所望の性質を満たす特徴量を設計するという枠組みを提案する．
-
Grant number：20H04157 2020.04 - 2023.03
Grant-in-Aid for Scientific Research Grant-in-Aid for Scientific Research(B)
Authorship：Coinvestigator(s) Grant type：Competitive
Grant amount：\900000 （ Direct Cost: \900000 ）
-
2017.04 - 2020.03
Grant-in-Aid for Scientific Research Grant-in-Aid for Scientific Research(B)
-
Systems Science of Bio-Navigation
2016.06 - 2021.03
Grant-in-Aid for Scientific Research Grant-in-Aid for Scientific Research on Innovative Areas (Research in a proposed research area)
-
ナビゲーションにおける画像情報分析基盤の整備とヒトの行動分類
2016.06 - 2021.03
科学研究費補助金 新学術領域研究(研究領域提案型)
玉木 徹、玉木 徹, 藤吉 弘亘
本研究では，本計画班の構成員が開発してきた最先端の映像認識技術に立脚し，野生動物やペットなどに装着したカメラから得られた映像や，人間が撮影した映 像など，これまでの映像認識技術では処理が困難な自己移動を含む映像を，安定かつ頑健に認識する技術を開発し，本領域における画像・映像情報分析のための 基盤技術を構築する．本年度の実績は以下のとおりである．
・前年度までに，B01生態学チームから提供された海鳥のGPU経路データを学習し，目的地までに至る経路を予測するための逆強化学習を利用した手法を開発している．これをさらに発展させて，GPS経路データの欠損部分を補完する手法を開発した．これにより，これまでは様々な原因で得られなかった経路情報が，データ駆動型モデルによりもっともらしい経路を出力することが可能になり，また補完経路を確率分布として出力することが可能となった．しかしこの手法は膨大な計算時間と多大なメモリ量を必要とするため，制度を保ちつつ計算コストを大幅に削減する手法を考案した．
・映像中の人物移動軌跡をいくつかのグループに分け（クラスタリングし），歩行目的地に応じて分割する手法を，さらに発展させた．これは前年度までに開発したベイズ推定に基づく手法である．それぞれの目的地へと到達する様子カーネル密度推定を用いて可視化し，どのような経路と目的地が頻繁に利用されているのかを把握することが可能となった．
・B01生態学チームから提供されたコウモリの音声データから3次元位置を予測する手法を開発した．屋内で飛行するコウモリの3次元位置を，20chのマイクロホンアレイで録音された音声信号から，回帰によって推定する深層ネットワークを提案し，20cm程度の誤差（RMSE）で推定することが可能となった．
Teaching Experience
-
科学技術計算
2022.04 Institution：Nagoya Institute of Technology
Level：Undergraduate (specialized) Country：Japan
-
画像処理特論IV
2021.10 Institution：Nagoya Institute of Technology
Level：Postgraduate Country：Japan
-
メディア系演習II
2021.10 Institution：Nagoya Institute of Technology
Level：Undergraduate (specialized) Country：Japan
-
プログラミング基礎
2021.10 - 2023.03 Institution：Nagoya Institute of Technology
Level：Undergraduate (specialized) Country：Japan
-
ソフトウェア工学
2021.04 Institution：Nagoya Institute of Technology
Level：Undergraduate (specialized) Country：Japan
Committee Memberships
-
電子情報通信学会 英文論文誌EＤ編集委員
2022.05 - 2024.05
Committee type：Academic society
-
電子情報通信学会 パターン認識・メディア理解研究専門委員会 専門委員
2020.06 - 2022.06
Committee type：Academic society
-
電子情報通信学会 パターン認識・メディア理解研究専門委員会 副委員長
2018.06 - 2020.06
Committee type：Academic society
-
情報処理学会 コンピュータビジョンとイメージメディア研究運営委員会 運営委員
2016.04 - 2020.03
Committee type：Academic society
-
電子情報通信学会 医用画像研究専門委員会 専門委員
2014.06 - 2022.06
Committee type：Academic society
-
情報処理学会 コンピュータグラフィックスとビジュアル情報学研究運営委員会 運営委員
2013.04 - 2017.03
Committee type：Academic society
-
電子情報通信学会 パターン認識・メディア理解研究専門委員会 専門委員
2012.05 - 2014.06
Committee type：Academic society
-
電子情報通信学会 パターン認識・メディア理解研究専門委員会 幹事
2011.05 - 2012.05
Committee type：Academic society
-
電子情報通信学会 ソサイエティ論文誌編集委員会 査読委員
2010.08
Committee type：Academic society
Social Activities
-
Role(s)： Presenter, Planner, Organizing member
connpass 2022.11
Audience： College students, Graduate students, Teachers, Researchesrs, General, Scientific, Company, Governmental agency
Type：Seminar, workshop
-
Role(s)： Presenter, Planner, Organizing member
connpass 2022.11
Audience： College students, Graduate students, Teachers, Researchesrs, General, Scientific, Company, Governmental agency
Type：Seminar, workshop
-
Role(s)： Presenter, Planner, Organizing member
connpass 2022.10
Audience： College students, Graduate students, Teachers, Researchesrs, General, Scientific, Company, Governmental agency
Type：Seminar, workshop
-
Role(s)： Presenter, Planner, Organizing member
connpass 2022.10
Audience： College students, Graduate students, Teachers, Researchesrs, General, Scientific, Company, Governmental agency
Type：Seminar, workshop
-
出張授業
Role(s)： Lecturer
愛知県立旭野高等学校 2022.07
-
Role(s)： Presenter, Planner, Organizing member
connpass 2022.06
Audience： College students, Graduate students, Teachers, Researchesrs, General, Scientific, Company, Governmental agency
Type：Seminar, workshop
-
Role(s)： Presenter, Planner, Organizing member
connpass 2022.04
Audience： College students, Graduate students, Teachers, Researchesrs, General, Scientific, Company, Governmental agency
Type：Seminar, workshop
-
公開講座：コンピュータサイエンス・アドベンチャー~理論計算機科学はこんなに面白い!~
Role(s)： Lecturer
名古屋工業大学 学務課 学務企画係 2021.11