プロフィール
自己紹介
2014年慶應義塾大学大学院理工学研究科後期博士課程修了,博士(工学).2013,2014年 ミュンヘン工科大学Visiting Scientist,2014年東京大学日本学術振興会特別研究員(PD),2015年産総研特別研究員を経て,2023年4月より国立研究開発法人 産業技術総合研究所 上級主任研究員.コンピュータビジョン分野および周辺分野の網羅的論文調査や連携研究を実施する研究コミュニティ cvpaper.challenge 主宰.時空間モデルのベースラインである3D ResNetの研究開発,実画像を用いない画像認識AIの事前学習法である数式ドリブン教師あり学習(Formula-Driven Supervised Learning; FDSL)を提案.2011/2020年ViEW小田原賞,2013年電気学会誌論文奨励賞,2014年藤原賞,2016年ECCV Workshop Brave New Idea,2019年度産総研論文賞,2020年 ACCV 2020 Best Paper Honorable Mention Award.研究はMIT Technology Reviewや日経ロボティクスなどのメディアにて掲載.
-
現職
オックスフォード大学 Visual Geometry Group(Oxford VGG)訪問研究員
(2024年9月 - ) -
最終学歴
慶應義塾大学大学院 理工学研究科 博士(工学)
(2011年4月 - 2014年3月)
プロジェクト
-
cvpaper.challenge
We are finding a collaborator to read/write a sophisticated paper!
-
Pre-training without Natural Images
We would like to replace Supervised/Self-supervised Learning!
-
Alleviating Over-segmentation Errors by Detecting Action Boundaries
Detecting action boundary significantly improves segmentation performance.
-
View-agnostic Image Rendering
We generate novel-view images.
-
Scene Change Captioning
We can describe a change area in a real environment.
-
Weakly Supervised Person Dataset (WSPD)
Our weak-supervision surpassed a supervised pre-training.
-
Neural Joking Machine: Humorous image captioning
Now we are joking!
-
Drive Video Analysis for the Detection of Traffic Near-Miss Incidents
We have collected large-scale traffic near-miss incident database!
-
Can Spatiotemporal 3D CNNs Retrace the History of 2D CNNs and ImageNet?
3D Conv is ready to be used various video applications!
-
Dynamic Fashion Cultures
Now we can start a world-wide fashion analysis!
主な論文
-
Selected Papers Top-Rank
Hirokatsu Kataoka, Ryo Hayamizu, Ryosuke Yamada, Kodai Nakashima, Sora Takashima, Xinyu Zhang, Edgar Josafat Martinez-Noriega, Nakamasa Inoue, Rio Yokota, “Replacing Labeled Real-Image Datasets with Auto-Generated Contours”, IEEE/CVF International Conference on Computer Vision and Pattern Recognition (CVPR), 2022.(Acceptance rate: 25.3%; 1st place in Computer Vision at Google Scholar Metrics)
-
Selected Papers Top-Rank
Ryosuke Yamada*, Hirokatsu Kataoka*, Naoya Chiba, Yukiyasu Domae Tetsuya Ogata, “Point Cloud Pre-training with Natural 3D Structures”, IEEE/CVF International Conference on Computer Vision and Pattern Recognition (CVPR), 2022.(Acceptance rate: 25.3%; 1st place in Computer Vision at Google Scholar Metrics; * indicates equal contribution)
-
Selected Papers Top-Rank
Hirokatsu Kataoka, Kazushige Okayasu, Asato Matsumoto, Eisuke Yamagata, Ryosuke Yamada, Nakamasa Inoue, Akio Nakamura, Yutaka Satoh, “Pre-training without Natural Images”, International Journal of Computer Vision (IJCV), 2022. (IF: 7.410)
-
Selected Papers Top-Rank
Kodai Nakashima, Hirokatsu Kataoka, Asato Matsumoto, Kenji Iwata, Nakamasa Inoue, Yutaka Satoh, “Can Vision Transformers Learn without Natural Images?,” AAAI Conference on Artificial Intelligence (AAAI), 2022.
-
Selected Papers Top-Rank
Hirokatsu Kataoka, Kazushige Okayasu, Asato Matsumoto, Eisuke Yamagata, Ryosuke Yamada, Nakamasa Inoue, Akio Nakamura, Yutaka Satoh, “Pre-training without Natural Images”, Asian Conference on Computer Vision (ACCV), 2020. (Best Paper Honorable Mention Award; Oral Presentation; 3 Strong Accepts)
研究チーム
合計16 名-
Yanjun Sun
( w/ Yue Qiu ) -
大谷 豪 / Go Ohtani
-
森 江梨花 / Erika Mori
( w/ Qiu Yue ) -
大塚 大地 / Daichi Otsuka
( w/ ALラボ )
-
岩片 彰吾 / Shogo Iwakata
( w/ Qiu Yue ) -
福沢 匠 / Takumi Fukuzawa
( w/ 原健翔 ) -
内山 史也 / Fumiya Uchiyama
( w/ Qiu Yue )