Explore the research
Building IoT Prediction with Sparse Graphs and Physical Constraints
Spatiotemporal prediction that limits graph neighbors and incorporates bandwidth and connection constraints into training.
Recommending Learning Actions from Activity Logs
A study of learning-action recommendations evaluated through changes in a grade predictor's outputs when activity logs are modified.
We study representations for real-world sensing data such as video, LiDAR, IMU, network logs, and IoT signals.
The goal is to capture temporal and spatial structure for action recognition, trajectory prediction, forecasting, anomaly detection, and infrastructure monitoring.
Why It Is Difficult
Real-world sensing data contains noise, missing values, occlusion, domain drift, and irregular sampling.
Spatiotemporal representations are high-dimensional, and many applications require efficient inference or online adaptation.
Approach
We use transformers, graph neural networks, self-supervised learning, multimodal fusion, and efficient inference methods.
For IoT and infrastructure problems, we also examine how physical constraints and graph sparsity can be integrated into learning.
Evaluation
We combine benchmark evaluation with real-environment validation.
Depending on the problem, we separately evaluate predictive accuracy, robustness to missing data, latency, computational complexity, and operational interpretability.
Current Questions
- 1Action summarization and anomaly-cue extraction from long videos
- 2Robust multimodal sensing through sensor fusion
- 3Adaptive spatiotemporal recognition with online updates
Related publications (13)
Dynamic BGP Route Control Based on Real-Time Performance Information Using an LLM with a Verification Mechanism
Combines LLM-based decisions with a verification mechanism to study network route control using real-time performance information.
Tsunogai, Kento, Yamasaki, Shugo, Kawabe, Kaito, Usami, Hiroyasu
Proceedings of the 2026 Tokai-Section Joint Conference on Electrical, Electronics, Information, and Related Engineering, 2026 (8)
2B-2-7, p. 154, Network Systems, August 30, 2026
Efficient 2D Human Pose Estimation via Token Compression and Cross-Attention Restoration
Studies efficient human pose estimation by compressing image tokens and reconstructing information through cross-attention.
Hashizume, Wataru, Usami, Hiroyasu
Proceedings of the 2026 Tokai-Section Joint Conference on Electrical, Electronics, Information, and Related Engineering, 2026 (8)
1C-3-3, p. 188, Computer Vision and Image Media, August 30, 2026
2D Human Pose Estimation via Compressed Token Learning with Structural Consistency
Studies human pose estimation with compressed image tokens while considering structural consistency.
Hashizume, Wataru, Usami, Hiroyasu
Meeting on Image Recognition and Understanding (MIRU 2026), 2026 (8)
IS3-171, Paper ID 366, Interactive Session IS3, 2026-08-06
O(V) Computational Optimization and Physics-informed STGNN for Building IoT Simulation
Studies building IoT prediction by combining graph sparsification for efficient computation with physical operational constraints.
Tajima, Koichiro, Usami, Hiroyasu
IEICE Technical Report, 2026 (1)
EST2025-95, 2026-01-23
Rain-effect Removal and Weather Estimation for Outdoor Images Using Rain Artifact Separation Learning
Studies separation of rain artifacts in outdoor images for image restoration and weather estimation.
Kikuta, Manato, Hashizume, Wataru, Usami, Hiroyasu
23rd Workshop on Informatics (WiNF 2025), 2025
P1A-31
Application of a Fully Connected Spatio-Temporal GNN Structure for Day-of-week Classification Based on Communication Logs
Studies weekday classification from communication logs using a fully connected spatiotemporal graph neural network.
Tajima, Koichiro, Nakashima, Yuki, Usami, Hiroyasu
Proceedings of the 2025 Tokai-Section Joint Conference on Electrical, Electronics, Information, and Related Engineering, 2025 (9)
G3-6, 2025-09-18
Recommending Learning Actions Using Neural Network
A study of learning-action recommendations evaluated through changes in a grade predictor's outputs when activity logs are modified.
Kohama, Hirokazu, Ban, Yuki, Hirakawa, Tsubasa, Yamashita, Takayoshi, Fujiyoshi, Hironobu, Itai, Akitoshi, Usami, Hiroyasu
Proceedings of the 31st International Conference on Computers in Education, 2023
ICCE 2023
Image Processing-based Support System for Recognizing Dangerous Worker Actions and Preventing Machine Accidents
Studies image-based recognition of hazardous worker actions as a basis for machinery-accident prevention support.
Hashizume, Wataru, Usami, Hiroyasu, Noguchi, Yusuke, Tsuchida, Hiroyuki, Iwahori, Yuji
21st Workshop on Informatics (WiNF 2023), 2023
S2A-1
Sample-region Extraction in Biomass Ash Dissolution Image Data Using Time-series Information
Studies region extraction from biomass-ash melting images using temporal information.
Yamane, Haruki, Usami, Hiroyasu, Suzuki, Yuri, Ninomiya, Yoshihiko
IEICE Technical Report, 2023
EST2022-87
Shape Classification of Fly Ash Particles Using Deep Learning
Studies classification of fly-ash particle shapes using deep learning.
Ninomiya, Yoshihiko, Jiang, Yanchi, Nakamura, Taichi, Usami, Hiroyasu, Mori, Takehito, Naganuma, Hiroshi, Sawada, Akihiro
Proceedings of the Conference on Coal Science, 2022
No. 59, pp. 20-21
A Study on Biomass Ash Melting Test Methods and Their Interpretation
Examines testing methods and interpretation of biomass-ash melting behavior.
Ninomiya, Yoshihiko, Sato, Ryuma, Usami, Hiroyasu, Hoshino, Hiroki
Proceedings of the Conference on Biomass Science, 2022
O-18
Adaptive Virtualization Environment Using Edge Computing
Studies the design of an adaptive virtualization environment using edge computing.
Hayashi, Yuto, Yamane, Haruki, Usami, Hiroyasu
20th Workshop on Informatics (WiNF 2022), 2022
S-1B-5
Coal Image Segmentation Using an Attention Mechanism Considering Relationships Between Coal and Included Minerals
Studies coal-image segmentation with attention to relationships between coal and embedded minerals.
Tajima, Koichiro, Doi, Mizuki, Usami, Hiroyasu, Ninomiya, Yoshihiko
20th Workshop on Informatics (WiNF 2022), 2022
S-4B-10