CSIRO researcher advancing radar and visual place recognition with learned, geometry-aware representations.
2
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-02) 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: no publication-volume data is available in the current feed. This axis would measure total and annual research output — preprints, journal and conference papers — for comparison against robotics and computer vision peers.
Impact—
Not yet scored: citation counts, h-index, and download statistics are absent from this profile's data. This axis would measure how widely the reviewed place-recognition papers are cited and adopted by other researchers.
Focus0/100
Scores 0/100 on field concentration. The raw field distribution splits evenly — 50% robotics / computer vision (radar place recognition) and 50% computer vision / robotics (visual place recognition). In substance this is one tightly held problem, place recognition, addressed through two sensor modalities; the zero score reflects formal field labeling rather than scattered interests.
Collaboration41/100
Scores 41/100 from 7 distinct coauthors across the cataloged papers. The recurring set — Peyman Moghadam, Sagun Singh Shrestha, Abdelwahed Khamis, Walter Nedov, Kavindie Katuwandeniya, David Hall, and Kaushik Roy — points to a compact, CSIRO-anchored team: Rahman leads one 2026 paper as first author and contributes as a middle author on the other.
Activity—
Not yet scored: the feed lacks year-by-year publication records beyond the two 2026 preprints, so publishing cadence cannot be computed. This axis would measure recent output rate and continuity of research engagement over time.
Reproducibility55/100
Scores 55/100 — the mean across 2 Evolve-assessed papers, each rated medium repro readiness. The radar paper's headline claim (MulRan R@1@5m 91.3 vs RadarLoc's 87.7) is checkable against public benchmarks, though the review flags untested SOTA baselines; the DAD paper is well ablated, but its WildCross gain (61.41→66.37 R@1) rests on a single benchmark.
Research summary
Saimunur Rahman is a researcher at the Commonwealth Scientific and Industrial Research Organisation (CSIRO) whose cataloged work centers on place recognition — enabling robots to recognize previously visited locations — across two sensor modalities. The Evolve catalog holds two 2026 arXiv preprints. The first, with Rahman as first author, introduces spatially gated feature-correlation (covariance pooling) representations for automotive mmWave spinning-radar place recognition; Evolve's review reports MulRan R@1@5m of 91.3 versus 87.7 for RadarLoc, plus zero-shot transfer to HeRCULES, while noting untested state-of-the-art baselines. The second, a middle-author collaboration, presents DAD, a geometry-conditioned visual place recognition method injecting geometric-foundation-model depth (Depth Anything V2) into vision-foundation-model tokens (DINOv2, SALAD), raising WildCross inter-sequence R@1 from 61.41 to 66.37; the review praises its ablations but flags the single-benchmark evaluation. Both papers carry medium reproducibility readiness, averaging 55/100 across the two assessments. Seven distinct coauthors recur — including Peyman Moghadam and Abdelwahed Khamis — indicating a CSIRO-anchored team. Output, impact, and activity metrics are not yet available, so this remains an early-stage bibliometric profile built from the reviewed papers alone.
Collaboration pattern
Rahman's cataloged work shows a compact, institution-centered collaboration pattern: 7 distinct coauthors, all recurring across the Evolve-reviewed 2026 preprints. The core group — Peyman Moghadam, Sagun Singh Shrestha, Abdelwahed Khamis, Walter Nedov, Kavindie Katuwandeniya, David Hall, and Kaushik Roy — is consistent with a CSIRO robotics lab team; Rahman took the first-author lead on the radar paper and a middle-author role on the visual one. The collaboration score of 41 reflects this small, stable team rather than a broad network.
Research keywords
Highlights
Gated covariance pooling pushes spinning-radar place recognition past RadarLoc
Depth-conditioned tokens lift visual place recognition on WildCross
A shared recipe: geometry injected into learned place descriptors
Across both cataloged papers the same design pattern appears — structural or geometric signals (second-order covariance statistics, foundation-model depth) folded into learned global descriptors — suggesting a deliberate research program in modality-agnostic place recognition at CSIRO.
Timeline
- 2026Publishes first-author arXiv preprint on automotive mmWave spinning-radar place recognition with spatially gated feature-correlation representation (Evolve-reviewed; MulRan R@1@5m 91.3 vs 87.7 for RadarLoc).
- 2026Co-authors geometry-conditioned visual place recognition preprint (DAD), reviewed by Evolve, with WildCross inter-sequence R@1 improving from 61.41 to 66.37.
- 2026Both papers are assessed at medium reproducibility readiness, averaging 55/100, consolidating a CSIRO-based place-recognition profile with seven recurring coauthors.
Papers in the Evolve catalog
Frequent co-authors
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 dimensions are computed from OpenAlex-linked records only; output, impact, and activity axes lack underlying data.
- Only two 2026 arXiv preprints reviewed by Evolve are covered; the peer-reviewed publication record is not represented.
- Profile name-resolved from public bibliographic data via ORCID 0000-0002-5250-5612; no country recorded in the source.
- Reproducibility score (mean 55/100) derives from just 2 assessed papers and may shift as more are reviewed.