Hiromu Taketsugu

Toyota Technological Institute · Japan

Unclaimed · AI-structured profileOpenAlex

Computer-vision researcher at Toyota Technological Institute focused on human pose, gait, and robot learning.

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-09-20) and cover the author's entire public record, not just the papers listed here. Reproducibility signal comes from Evolve's own reviews.

Research profile

Output—

Not yet scored: this axis measures total publication volume and productivity trajectory, but the underlying output data for this profile is insufficient to compute a reliable score; the current record shows 7 works in the last three years.

Impact—

Not yet scored: this axis measures citation influence such as total citations, field-weighted impact, and h-index growth; beyond an h-index of 2 and an i10-index of 0, no citation data is available, so no reliable estimate is possible.

Focus100/100

Exceptionally concentrated (100/100). Of 19 topic-tagged works in the OpenAlex record, 13 — roughly 68% — fall under Human Pose and Action Recognition (9) and Gait Recognition and Analysis (4), with only thin coverage of surveillance and tracking (2), robot manipulation and learning (2), and autonomous vehicles (2). The two 2026 catalog papers extend this core into robot learning while remaining anchored in motion understanding.

Collaboration36/100

Modest collaboration footprint (36/100): 5 distinct co-authors in the record, with the frequent-coauthor list resolving to Norimichi Ukita, Toshiki Otani, and Bo-Han Chen after removing a self-entry and a duplicate. Both 2026 catalog papers are middle-author, non-corresponding contributions, consistent with a supporting role in small teams rather than leading large collaborations.

Activity61/100

Above-median activity (61/100): countsByYear shows 2 works in 2023, 1 in 2024, 3 in 2025, and 3 in 2026 — 7 works in the most recent three years, a steady cadence that rebounded after a quiet 2024. Recent output is preprint-forward: both catalog papers are 2026 arXiv submissions.

Reproducibility20/100

Low assessed reproducibility (20/100), based on a single assessed paper (reproMean 20). Evolve's review of the energy-regularized imitation learning work flags a claimed 2.1% work reduction that is small, task-concentrated, and reported without variance or significance analysis, with low reproducibility readiness; the PRISM paper has not yet been assessed.

Research summary

Hiromu Taketsugu is a computer-vision researcher at Toyota Technological Institute (Japan) whose public record, resolved from OpenAlex, centers on human motion understanding: Human Pose and Action Recognition accounts for 9 works and Gait Recognition and Analysis for 4, with thinner activity in video surveillance and tracking (2), robot manipulation and learning (2), and autonomous-vehicle technology and safety (2). The profile is early-stage: h-index 2, i10-index 0, and 9 works across 2023–2026 (2, 1, 3, 3 by year), 7 of them within the last three years — a steady recent cadence that rebounded after a quiet 2024. Focus is the strongest dimension (100/100), reflecting that roughly 68% of topic-tagged works sit in pose and gait research. Collaboration is modest (36/100) with 5 distinct co-authors, including Norimichi Ukita, Toshiki Otani, and Bo-Han Chen. The Evolve catalog holds two 2026 arXiv preprints, both middle-author and non-corresponding: the reviewed energy-regularized imitation learning paper fine-tunes RVT-2 with a differentiable joint-work predictor, and Evolve judges its claimed 2.1% work reduction small, task-concentrated, and unaccompanied by variance or significance analysis (reproducibility 20/100, from one assessed paper). PRISM, on social robot navigation, is not yet reviewed.

Collaboration pattern

The record shows a small, stable collaboration network: 5 distinct co-authors overall (collaboration score 36/100), with the frequent-coauthor list resolving to three names after removing a self-entry and a duplicate — Norimichi Ukita, Toshiki Otani, and Bo-Han Chen. Both 2026 catalog papers list the researcher in a middle author position without a corresponding role, suggesting contributions within existing teams rather than independently led projects.

Research keywords

Human pose estimationAction recognitionGait recognitionGait analysisVideo surveillanceVisual trackingImitation learningRobotic manipulationEnergy-aware policy learningSocial robot navigationAutonomous vehicle safety

Highlights

Energy-aware fine-tuning of RVT-2 for lower-effort manipulation
This reviewed 2026 arXiv preprint trains a differentiable joint-work predictor to fine-tune RVT-2 so that manipulation policies reduce force and mechanical work in simulation. Evolve judges the claimed 2.1% work reduction small, task-concentrated, and reported without variance or significance analysis, and rates reproducibility readiness as low.
PRISM: predictive representations for social robot navigation
A second 2026 arXiv preprint proposes PRISM, a predictive representation of interaction style and motion for social robot navigation. It is a middle-author, non-corresponding contribution and has not yet been reviewed by Evolve, so no verdict or reproducibility assessment is available.
Pose- and gait-centred research profile
13 of the record's 19 topic-tagged works — about 68% — sit in Human Pose and Action Recognition (9) and Gait Recognition and Analysis (4), with smaller clusters in surveillance and tracking, robot manipulation, and autonomous vehicles, producing the maximum focus score of 100/100.

Timeline

2023Earliest works in the OpenAlex record appear: 2 publications within the pose and gait core.
2024Quietest year in the record, with 1 publication.
2025Output rebounds to 3 publications.
2026Most active year: 3 publications, including two arXiv preprints in the Evolve catalog — the energy-regularized imitation learning paper and PRISM for social robot navigation.

Papers in the Evolve catalog

Frequent co-authors

Bo-Han Chen×1 Hiromu Taketsugu×1 Norimichi Ukita×1 Toshiki Otani×1 Norimichi Ukita×1

About this profile

This profile is AI-structured by Evolve.Pub from public bibliographic data (OpenAlex) and Evolve's own paper reviews. It is not written or endorsed by the researcher unless the profile is claimed. Radar dimensions are computed from real bibliometrics with fixed formulas — raw figures are shown alongside; missing dimensions mean insufficient data, never zero. Errors? Claim the profile to correct or hide it.

Bibliometric figures (h-index 2, i10-index 0, counts by year, topics, co-authors) come from an OpenAlex-resolved identity only; no ORCID was provided, so name resolution may be imperfect for this name.
Output and impact dimensions are unscored because the underlying data is insufficient; no citation data beyond the h-index and i10-index is available.
Only one catalog paper has been reviewed by Evolve; the PRISM paper lacks a verdict, keywords, and a reproducibility assessment.
The frequent-coauthor list contained a self-entry and a duplicate name; the collaboration analysis uses the deduplicated set.