Biography
I am Ryo Nakamura, working as a data scientist at Tenchijin Inc. In my work, I tackle social problems by using Earth-related data to reduce costs and automate operations.
Additionally, as personal activities, I organize the SatAI.challenge (a community for learning satellite data and AI) and cvpaper.challenge. Through these, I aim to boost the growth of satellite data, AI, and computer vision technologies in Japan.
Under the guidance of Professor Atsushi Fujiki (formerly Professor Masaru Tanaka) at Fukuoka University and working as a research assistant under Hirokatsu Kataoka at the National Institute of Advanced Industrial Science and Technology (AIST), I gained experience in having papers accepted at top computer vision conferences and organizing workshops. I obtained my Doctor of Science degree in 2023.
- Machine Learning
- Computer Vision
- Information Geometry
- Remote Sensing
Doctor of Philosophy (Mathematics of Information), 2024
Fukuoka University
Master of Science (Mathematics of Information), 2021
Fukuoka University
Bachelor of Science (Mathematical Informatics in Society), 2019
Fukuoka University
Experience
Archiving Japanese Materials (You can check the materials on the following links)
Research
- Formula-Driven Supervised Learning
- Traffic Accident Video Recognition
- Out of distribution detection with q-Gaussian distribution
Work
- Research Efficiency Tips
- High-impact Papers in Computer Vision
- cvpaper.challenge conference 2021 (CCC2021) Program Chair
- cvpaper.challenge conference winter 2021 (CCCW2021) General Chair
- ViEW2021 Invited Talk Tutorial Session “How to Start Latest Research”
- Research Efficiency Tips ver2
- Meta Survey: Formula-Driven Supervised Learning
- Meta Survey: Foundation Models
- CVPR2022 Comprehensive Survey
- cvpaper.challenge conference summer 2022 (CCCS2022) Program Chair
- ECCV2022 Comprehensive Survey
- CVPR2023 Quick Report
Work
- Research and Development for Expanding Radio Resources
- Satellite Image Analysis
- Latest Technology Research and Review
- Data Science Team Leader
Achievement
Achievement
・Simple variational inference based on minimizing Kullback–Leibler divergence
Ryo Nakamura 1, Tomooki Yuasa 2, Takafumi Amaba 1, Jun Fujiki 1 (1. Fukuoka Univ., 2. Faculty of Economics and Business Administration), Information Geometry
・Scaling Backwards: Minimal Synthetic Pre-training?
Ryo Nakamura 1 *, Ryu Tadokoro 1 *, Ryosuke Yamada 1, Yuki M. Asano 2, Iro Laina 3, Christian Rupprecht 3, Nakamasa Inoue 4, Rio Yokota 4, Hirokatsu Kataoka 1 (1. AIST, 2. University of Amsterdam, 3. University of Oxford, 4. Tokyo Institute of Technology, * These authors contributed equally), European Conference on Computer Vision 2024 (ECCV 2024)
・On the Relationship Between Double Descent of CNNs and Shape/Texture Bias Under Learning Process
Shun Iwase 1, Eisaku Maeda 1, Nakamasa Inoue 2, Hirokatsu Kataoka 3, Ryo Nakamura 3, Rio Yokota 2, Shuya Takahashi 1 (1. Tokyo Denki Univ., 2. Tokyo Institute of Technology, 3. AIST), International Conference on Pattern Recognition 2024 (ICPR 2024)
・Pseudo-outlier synthesis using q-Gaussian distributions for out-of-distribution detection
Ryo Nakamura 1,2, Ryu Tadokoro 3, Eisuke Yamagata 4, Yusuke Kondo 5, Kensho Hara 1, Hirokatsu Kataoka 1, Nakamasa Inoue 3 (1. AIST, 2. Fukuoka Univ., 3. Tohoku Univ., 4. Tokyo Univ., 5. Tokyo Institute of Technology), International Conference on Acoustics, Speech, and Signal Processing 2023 (ICASSP 2023)
・Primitive Geometry Segment Pre-training for 3D Medical Image Segmentation
Ryu Tadokoro1,2, Ryosuke Yamada 2,3, Kodai Nakashima 2,3, Ryo Nakamura2, Hirokatsu Kataoka 2 (1. Tohoku Univ., 2. AIST, 3. Tsukuba Univ.), British Machine Vision Conference 2023 (BMVC 2023)
・Pre-training Vision Transformers with Very Limited Synthesized Images
Ryo Nakamura1,2, Hirokatsu Kataoka 1, Sora Takashima 3, MARTINEZ-NORIEGA Edgar Josafat 3, Rio Yokota 3, Nakamasa Inoue3 (1. AIST, 2. Fukuoka Univ., 3. Tokyo Institute of Technology), International Conference on Computer Vision 2023 (ICCV 2023)
・Traffic Incident Database with Multiple Labels Including Various Perspective Environmental Information
Shota Nishiyama, Takuma Saito, Ryo Nakamura, Go Ohtani, Hirokatsu Kataoka, Kensho Hara (AIST), IEEE/RSJ International Conference on Intelligent Robots and Systems 2023 (IROS 2023)
・Rough Target Region Extraction with Background Learning
Ryo Nakamura, Yoshiaki Ueda, Masaru Tanaka, Jun Fijiki (Fukuoka University), International Workshop on Frontiers of Computer Vision 2023 (IW-FCV2023), 2023/2・IEEE/CVF International Conference on Computer Vision and Pattern Recognition (CVPR2021)
Hirokatsu Kataoka, Kensho Hara, Ryota Suzuki, Yoshihiro Fukuhara, Hideki Tsunashima, Kodai Nakashima, Ryo Nakamura, Ryosuke Yamada, Kai Watanabe (AIST), 2021/11/7・Output augmentation works well without any domain knowledge
Shu Eguchi, Ryo Nakamura, Masaru Tanaka (Fukuoka University), International Conference on Machine Vision Applications (MVA2021), 2021/7/25-27・Research and Development on Ka-band Satellite Control for Diverse Use Cases - Communication Demand and Line Condition Prediction Technology
Ayako Kimura, Atsushi Ueda, Yutaka Kawashima, Qu Wanding, Shunsuke Kogure, Kenichi Sasaki, Sarah Nakama Yoko, Ryo Nakamura, Hideto Horie (Tenchijin Inc.), 2021 IEICE General Conference, 2021/3/9・Target Region Extraction Using Asymmetry of Training Image Set
Ryo Nakamura, Yoshiaki Ueda, Jun Fujiki, Masaru Tanaka (Fukuoka Univ.), Pattern Recognition and Media Understanding (PRMU), 2019/3/17・Aperture Synthesis Using CNN - Toward Reducing Annotation Work
Ryo Nakamura, Masaru Tanaka, Jun Fujiki, Yoshiaki Ueda (Fukuoka Univ.), Neurocomputing Research Group (NC), 2019/12/06Presentation
・Pre-training Vision Transformers with Limited Synthesized Images
Ryo Nakamura1,2, Hirokatsu Kataoka1, Sora Takashima3, MARTINEZ-NORIEGA Edgar Josafat3, Rio Yokota3, Nakamasa Inoue3 (1. National Institute of Advanced Industrial Science and Technology, 2. Fukuoka University, 3. Tokyo Institute of Technology), 29th Symposium on Sensing via Image Information (SSII2023), 2023/6/14, Interactive Session Audience Award
・Can Visual Features Be Learned with Only One Generated Image Per Category?
Ryo Nakamura (Fukuoka Univ., AIST), Ryu Tadokoro (Tohoku Univ.), Hirokatsu Kataoka (AIST), Meeting on Image Recognition and Understanding (MIRU2022), 2022/7/26
・How to Start Latest Research
Ryo Nakamura (Fukuoka Univ.), Vision Engineering Workshop 2021 (ViEW2021), Tutorial Session, 2021/12/2
・Rough Target Region Extraction Using VAE with Background Learning
Ryo Nakamura, Yoshiaki Ueda, Jun Fujiki, Masaru Tanaka (Fukuoka Univ.), Information-Based Induction Sciences 2020 (IBIS2020), 2020/11/26
・Target Region Extraction by Background Learning with Filter Size Selection Mechanism
Yoshiaki Ueda, Ryo Nakamura, Jun Fujiki, Masaru Tanaka (Fukuoka Univ.), Information-Based Induction Sciences 2020 (IBIS2020), 2020/11/26
・Target Region Extraction by Background Learning
Ryo Nakamura, Yoshiaki Ueda, Jun Fujiki, Masaru Tanaka (Fukuoka Univ.), Meeting on Image Recognition and Understanding (MIRU2020), 2020/08/02-05, Student Encouragement Award
・CNN with Aperture Synthesis
Ryo Nakamura, Yoshiaki Ueda, Masaru Tanaka (Fukuoka Univ.), Information-Based Induction Sciences 2019 (IBIS2019), 2019/11/20