Tatsunori Taniai, Ph.D.
谷合竜典

I am a Senior Researcher in the Knowledge Computing Group at OMRON SINIC X Corporation, based in Tokyo, Japan. Through my research, I am pursuing a new methodology for the era of deep learning, which I call "Principled AI." My research interests span sensing, 3D vision, atomic system modeling, and autonomy——areas where physical, mechanical, or algorithmic principles play a fundamental role.

Topics: Sensing, 3D Vision, Physics-Informed ML, Autonomy.

Call for Interns: I am seeking an intern to work intensively with me on a geometry-related problem in 3D computer vision or physics-informed machine learning, with the goal of publishing a paper at CVPR, ICCV, ECCV, ICML, NeurIPS, or ICLR.

Email: tatsunori.taniai [at] sinicx.com

Research Projects

Rethinking the role of frames for SE(3)-invariant crystal structure modeling

ICLR 2025

Yusei Ito*, Tatsunori Taniai*, Ryo Igarashi, Yoshitaka Ushiku, Kanta Ono

Paper / Project / Code
Crystalformer: Infinitely Connected Attention for Periodic Structure Encoding

ICLR 2024

Tatsunori Taniai, Ryo Igarashi, Yuta Suzuki, Naoya Chiba, Kotaro Saito, Yoshitaka Ushiku, Kanta Ono

Paper / Project / Code
Neural structure fields with application to crystal structure autoencoders

Communication Materials 2023

Naoya Chiba, Yuta Suzuki, Tatsunori Taniai, Ryo Igarashi, Kotaro Saito, Yoshitaka Ushiku, Kanta Ono

Open Access / Project / Code
Risk-aware Path Planning via Probabilistic Fusion of Traversability Prediction for Planetary Rovers on Heterogeneous Terrains

ICRA 2023

Masafumi Endo, Tatsunori Taniai, Ryo Yonetani, Genya Ishigami

Preprint / Project / Code
Self-supervised learning of materials concepts from crystal structures via deep neural networks

Machine Learning: Science and Technology 2022

Yuta Suzuki, Tatsunori Taniai, Kotaro Saito, Yoshitaka Ushiku, Kanta Ono

Open Access / Code
Quasistatic contact-rich manipulation via linear complementarity quadratic programming

IROS 2022

Sotaro Katayama, Tatsunori Taniai, Kazutoshi Tanaka

Preprint / Project / Code
Path Planning using Neural A* Search

ICML 2021

Ryo Yonetani*, Tatsunori Taniai*, Mohammadamin Barekatain, Mai Nishimura, Asako Kanezaki

Preprint / Project / Code
Neural Inverse Rendering for General Reflectance Photometric Stereo

ICML 2018

Tatsunori Taniai, Takanori Maehara

Preprint / Project / Code
Continuous 3D Label Stereo Matching using Local Expansion Moves

TPAMI 2018

Tatsunori Taniai, Yasuyuki Matsushita, Yoichi Sato, Takeshi Naemura

Preprint / Project / Code
Semi-Global Stereo Matching with Surface Orientation Priors

3DV 2017 Spotlight

Daniel Scharstein, Tatsunori Taniai, Sudipta N. Sinha

Preprint (extended) / Poster / Slides
Fast Multi-frame Stereo Scene Flow with Motion Segmentation

CVPR 2017

Tatsunori Taniai, Sudipta N. Sinha, Yoichi Sato

Preprint / Preprint (extended) / Poster / Project / Demo
Joint Recovery of Dense Correspondence and Cosegmentation in Two Images

CVPR 2016

Tatsunori Taniai, Sudipta N. Sinha, Yoichi Sato

Preprint / Preprint (extended) / Poster / Project / Demo
Superdifferential Cuts for Binary Energies

CVPR 2015

Tatsunori Taniai, Yasuyuki Matsushita, Takeshi Naemura

Preprint / ExtAbst / Poster / Project / Code
Graph Cut based Continuous Stereo Matching using Locally Shared Labels

CVPR 2014

Tatsunori Taniai, Yasuyuki Matsushita, Takeshi Naemura

Preprint / Poster / Project
Image Segmentation using Dual Distribution Matching

BMVC 2012 Oral

Tatsunori Taniai, Viet-Quoc Pham, Keita Takahashi, Takeshi Naemura

Preprint / Supp / ExtAbst / Slides / VideoLecuture / Project

Experiences

Education

Ph.D. in Information Science and Technology, The University of Tokyo, 2017. Supervised by Prof. Yoichi Sato
M.Sc. in Information Science and Technology, The University of Tokyo, 2014. Supervised by Prof. Takeshi Naemura
B.Eng. in Information and Communication Engineering, The University of Tokyo, 2012. Supervised by Prof. Takeshi Naemura

Professional Experience

Senior Researcher at OMRON SINIC X Corporation, Tokyo, Japan, 2019 – Present
Visiting Researcher at RIKEN AIP Center, Tokyo, Japan, 2019 – 2020
Special Postdoctoral Researcher at RIKEN AIP Center, Tokyo, Japan, 2017 – 2019
Research Assistant at RIKEN AIP Center, Tokyo, Japan, 2017
JSPS Young Research Fellow at The University of Tokyo, Tokyo, Japan, 2014 – 2017

Internships

Research Intern at Microsoft Research, Redmond, USA, May – Aug. 2016. Supervised by Dr. Sudipta Sinha. Published a CVPR 2017 paper.
Visiting Researcher at Microsoft Research Asia, Beijing, China, Jan. – Apr. 2016. Supervised by Dr. David Wipf
Research Intern at Microsoft Research, Redmond, USA, May – Aug. 2016. Supervised by Dr. Sudipta Sinha. Published a CVPR 2016 paper.
Research Intern at Microsoft Research Asia, Beijing, China, Dec 2012 – Apr. 2013. Supervised by Dr. Yasuyuki Matsushita. Published a CVPR 2014 paper.

Service

Conference Reviewer: CVPR '25 – '22, '20, '18, ICCV '23, '19, '17, ECCV '20, 3DV '18, '14
Journal Reviewer: TPAMI '23, '19, '18, IJCV '18, CVIU '17, IMAVIS '16, IEEE TIP '18, '15.
SSII 2025, An Organizer of Tutorial and Technical Talk Sessions.

Awards and Honors

Outstanding Reviewer Recognitions from CVPR 2023 , ECCV 2020 , CVPR 2020 , CVPR 2018.
IPSJ Yamashita SIG Research Award (山下記念研究賞) from Information Processing Society of Japan, 2018 [Link]
Dean's Award for Best Doctoral Thesis from Graduate School of Information Science and Technology, The University of Tokyo, 2017
Microsoft Research Asia Fellowship 2015 from Microsoft Research Asia for thirteen PhD students from Asia majoring in CS, 2015 – 2017
JASSO Scholarship for Top 10% Excellent Master Students from Japan Student Services Organization, 2014
Dean's Award for Best Master Thesis from Graduate School of Information Science and Technology, The University of Tokyo, 2017
Dean's Award for Best Bachelor Thesis from Faculty of Engineering, The University of Tokyo, 2012

Publications

Books and Chapters

Tatsunori Taniai. Binocular Stereo. In Ikeuchi K. (eds) Computer Vision. Springer, Cham, 2020. [Preprint]

Journal Papers

Naoya Chiba, Yuta Suzuki, Tatsunori Taniai, Ryo Igarashi, Kotaro Saito, Yoshitaka Ushiku, Kanta Ono. Neural structure fields with application to crystal structure autoencoders. Communication Materials, 2023. [Open Access / Project / Code]
Yuta Suzuki, Tatsunori Taniai, Kotaro Saito, Yoshitaka Ushiku, Kanta Ono. Self-supervised learning of materials concepts from crystal structures via deep neural networks. Machine Learning: Science and Technology, 2022. [Open Access / Code]
Tatsunori Taniai, Yasuyuki Matsushita, Yoichi Sato, Takeshi Naemura. Continuous 3D Label Stereo Matching using Local Expansion Moves. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI 2018), vol. 40, no. 11, pp. 2725--2739, 2018. [Preprint / Project / Code]
谷合竜典, ファン・ヴェト・クォク, 高橋桂太, 苗村健. 前景・背景色分布の同時マッチングによる画像セグメンテーション. 電子情報通信学会和文論文誌D, 画像の認識・理解特集号, vol. J96-D, no. 8, pp. 1764--1777, 2013. (an extended version of our BMVC 2012's paper)

International Conference Papers

Yusei Ito*, Tatsunori Taniai*, Ryo Igarashi, Yoshitaka Ushiku, Kanta Ono. Rethinking the role of frames for SE(3)-invariant crystal structure modeling. In Proceedings of The Thirteenth International Conference on Learning Representations (ICLR 2025), Singapore, 2025. (acceptance rate 32%) [Paper / Project / Code]
Tatsunori Taniai, Ryo Igarashi, Yuta Suzuki, Naoya Chiba, Kotaro Saito, Yoshitaka Ushiku, Kanta Ono. Crystalformer: Infinitely Connected Attention for Periodic Structure Encoding. In Proceedings of The Twelfth International Conference on Learning Representations (ICLR 2024), Vienna, Austria, 2024. (acceptance rate 31%) [Paper / Project / Code]
Masafumi Endo, Tatsunori Taniai, Ryo Yonetani, Genya Ishigami. Risk-aware Path Planning via Probabilistic Fusion of Traversability Prediction for Planetary Rovers on Heterogeneous Terrains. In Proceedings of the 2023 IEEE International Conference on Robotics and Automation (ICRA 2023), pp. 11852--11858, London, UK, 2023. [Preprint / Project / Code]
Sotaro Katayama, Tatsunori Taniai, Kazutoshi Tanaka. Quasistatic contact-rich manipulation via linear complementarity quadratic programming. In Proceedings of the 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2022), pp. 203–-210, Kyoto, Japan, 2022. [Preprint / Project / Code]
Ryo Yonetani*, Tatsunori Taniai*, Mohammadamin Barekatain, Mai Nishimura, Asako Kanezaki. Path Planning using Neural A* Search. In Proceedings of the International Conference on Machine Learning (ICML 2021), a virtual conference, 2021. (acceptance rate 21%) [Preprint / Project / Code]
Tatsunori Taniai, Takanori Maehara. Neural Inverse Rendering for General Reflectance Photometric Stereo. In Proceedings of the International Conference on Machine Learning (ICML 2018), pp. 4864--4873, Stockholm, Sweden, 2018. (acceptance rate 25%) [Preprint / Project / Code]
Daniel Scharstein, Tatsunori Taniai, Sudipta N. Sinha. Semi-Global Stereo Matching with Surface Orientation Priors. In Proceedings of the 5th International Conference on 3D Vision (3DV 2017 Spotlight), pp. 215--224, Qingdao, China, 2017. [Preprint (extended) / Poster / Slides]
Tatsunori Taniai, Sudipta N. Sinha, Yoichi Sato. Fast Multi-frame Stereo Scene Flow with Motion Segmentation. In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2017), pp. 6891--6900, Honolulu, Hawaii, USA, 2017. (acceptance rate 30%) [Preprint / Preprint (extended) / Poster / Project]
Tatsunori Taniai, Sudipta N. Sinha, Yoichi Sato. Joint Recovery of Dense Correspondence and Cosegmentation in Two Images. In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2016), pp. 4246--4255, Las Vegas, NV, USA, 2016. (acceptance rate 30%) [Preprint / Preprint (extended) / Poster / Project]
Tatsunori Taniai, Yasuyuki Matsushita, Takeshi Naemura. Superdifferential Cuts for Binary Energies. In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2015), pp. 2030--2038, Boston, MA, USA, 2015. (acceptance rate 28%) [Preprint / ExtAbst / Project]
Tatsunori Taniai, Yasuyuki Matsushita, Takeshi Naemura. Graph Cut based Continuous Stereo Matching using Locally Shared Labels. In Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2014), pp. 1613--1620, Columbus, OH, USA, 2014. (acceptance rate 30%) [Preprint / Poster / Project / Code]
Tatsunori Taniai, Viet-Quoc Pham, Keita Takahashi, Takeshi Naemura. Image Segmentation using Dual Distribution Matching. In Proceedings of the 23rd British Machine Vision Conference (BMVC 2012), pp. 74.1--74.11, Surrey, UK, 2012. (oral presentation; acceptance rate 8%) [Preprint / Supp / ExtAbst / Slides / VideoLecuture / Project]

Invited Talks

Tatsunori Taniai, Sudipta N. Sinha, Yoichi Sato. Fast Multi-frame Stereo Scene Flow with Motion Segmentation (CVPR 2017). The 20th Meeting on Image Recognition and Understanding (MIRU 2017), Hiroshima, Japan, August 10th, 2017.
Tatsunori Taniai. Joint Recovery of Dense Correspondence and Cosegmentation in Two Images. The Workshop on Vision, Learning, and Cognition in Microsoft Research Asia Ph.D. Forum 2016, Microsoft office, Beijing, China, September 20th, 2016.
Tatsunori Taniai, Sudipta N. Sinha, Yoichi Sato. Joint Recovery of Dense Correspondence and Cosegmentation in Two Images (CVPR 2016). The 19th Meeting on Image Recognition and Understanding (MIRU 2016), IS2-15, Shizuoka, Japan, August 4th, 2016.
Tatsunori Taniai. Solving Segmentation and Dense Correspondence Problems using Graph Cuts (画像領域・対応点推定問題へのグラフカットの適用). The 1st CREST Symposium on Random Fields and Deep Learning, Waseda Univ., Tokyo, Japan, January 13th, 2016. (Organizers: Prof. Hiroshi Ishikawa & Prof. Takayuki Okatani) [Slides]
Tatsunori Taniai, Yasuyuki Matsushita, Takeshi Naemura. Superdifferential Cuts for Binary Energies (CVPR 2015). The 18th Meeting on Image Recognition and Understanding (MIRU 2015), IS1-10, Osaka, Japan, July 28th, 2015.
Tatsunori Taniai, Yasuyuki Matsushita, Takeshi Naemura. Graph Cut based Continuous Stereo Matching using Locally Shared Labels (CVPR 2014). The 17th Meeting on Image Recognition and Understanding (MIRU 2014), IT1-1, Okayama, Japan, July 29th, 2014.

Domestic Conference Papers

谷合竜典, 苗村健. 画素位置を埋め込んだ双色分布マッチングによる画像セグメンテーション. 画像の認識・理解シンポジウム（MIRU2013）予稿集, SS3-31, 東京, 2013.
谷合竜典, ファン・ヴェト・クォク, 高橋桂太, 苗村健. 前景および背景色分布の同時マッチングによる画像セグメンテーション. 画像の認識・理解シンポジウム（MIRU2012）予稿集, OS14-02, 福岡, 2012.

Thesis

Tatsunori Taniai. Discrete Inference Approaches to Image Segmentation and Dense Correspondence. Doctoral Thesis. March 2017. The University of Tokyo. Adviser: Yoichi Sato. [PDF / UT-Repo]
Tatsunori Taniai. Applying Graph Cuts to MAP Estimation of Continuous and Higher-Order Markov Random Fields. Master's Thesis. March 2014. The University of Tokyo. Adviser: Takeshi Naemura. [PDF]
Tatsunori Taniai. 前景・背景の大域的な色分布の同時マッチングによる画像セグメンテーション(Foreground Background Image Segmentation using Dual Distribution Matching). Bachelor's Thesis. March 2012. The University of Tokyo. Adviser: Takeshi Naemura. [PDF / Note]

Tatsunori Taniai, Ph.D.谷合 竜典