2D Human Pose Estimation via Compressed Token Learning with Structural Consistency
Hashizume, Wataru, Usami, Hiroyasu
Meeting on Image Recognition and Understanding (MIRU 2026), 2026
IS3-171, Paper ID 366, Interactive Session IS3, 2026-08-06
Department of Computer Science, College of Engineering, Chubu University
Understanding the world from images
We study image processing, computer vision, and machine learning, with applications in medicine, manufacturing, social infrastructure, and real-world sensing.
Our work connects principles and applications: from physics- and mathematics-based image understanding to systems for medical, industrial, and social problems.
Research on lesion detection, segmentation, 3D shape reconstruction, quantitative measurement, and diagnostic support using endoscopy, CT, cellular, and pathological images.
Research on estimating shape, material, reflectance, and illumination from images by modeling reflection, transmission, scattering, and other optical phenomena.
Research on extracting temporal and spatial structure from videos and sensor data for action recognition, forecasting, and anomaly detection.
Research on generative models, learning support, data augmentation, automatic question generation, and validation of generated outputs and LLM evaluators.
Hashizume, Wataru, Usami, Hiroyasu
Meeting on Image Recognition and Understanding (MIRU 2026), 2026
IS3-171, Paper ID 366, Interactive Session IS3, 2026-08-06
Usami, Hiroyasu, Hara, Keisuke, Tsuboi, Ayato, Matsuda, Naohiko
arXiv, 2026
arXiv:2606.15610 [cs.CL], 22 pages, 4 figures
Usami, Hiroyasu
Journal of Comprehensive Engineering, 2026
Vol. 38, pp. 1-11
Recent presentations, publications, awards, invited talks, academic service, outreach, and laboratory activities.
2026.08.06
Research by Wataru Hashizume and Hiroyasu Usami on 2D human pose estimation via compressed token learning with structural consistency is listed as an interactive presentation at MIRU 2026. The presentation is scheduled for Interactive Session IS3 on August 6, 13:00–15:00.
2026.06.14
The preprint “LLM Judges Have Dark Current: A Psychometric Datasheet for LLM-as-a-Judge Evaluation” is now available on arXiv. The work treats LLM-as-a-judge systems as measurement instruments rather than scalar accuracy or agreement devices, and proposes a Judge Datasheet protocol for measuring dark current, positional bias, target sensitivity, and criterion shift.
2026.06.04
A paper on an AI arteriosclerosis diagnosis system using RLHF and adversarial attack was published in Vol. 38 of Journal of Comprehensive Engineering. The study evaluates the robustness of IMT extraction from carotid ultrasound images under adversarial perturbations, focusing on whether attention remains on physician-relevant structures.
Faculty, graduate students, and undergraduate students work together across computer vision, machine learning, and applied image analysis.
1
Faculty
25
Current Students
21
Alumni
Companies are listed in Japanese alphabetical order.
Graduate study: Chubu University Graduate School
Students interested in image processing, computer vision, machine learning, and AI are welcome to contact us.
We welcome collaboration with companies and research institutions on image analysis, computer vision, and machine learning.