Georgia Tech researcher: geometry-grounded robot perception for recognition of visually similar industrial parts
2
papers here
2
reviewed by Evolve
1
first-author
0
corresponding
—
works · OpenAlex
2
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-16) 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—
Output cannot be scored yet: OpenAlex reports only minimal indicators (h-index 2, i10-index 0) for this early-stage profile. With fuller publication records, this axis would measure annual output volume and productivity trends across the researcher's career.
Impact—
Impact is unscored because citation evidence is too thin — i10-index 0 and h-index 2. Once citations accumulate, this axis would measure citation counts, h-index growth, and how far the industrial-perception methods are adopted.
Focus0/100
The radar focus score is 0, the lowest on the scale. OpenAlex splits this profile evenly (0.5/0.5) between two near-identical computer-vision labels and spreads work across five topic clusters — Robot Manipulation and Learning (4 works) plus four topics at 3 each. In practice, the two reviewed papers address one applied problem — model-free recognition of visually similar industrial parts — so the low score reflects taxonomy fragmentation as much as genuine dispersion.
Collaboration28/100
Collaboration scores 28. In the last three years OpenAlex records 13 works with only 3 distinct co-authors, and repeat collaboration centers on Alankrit Gupta and Seung-Kyum Choi. The reviewed 2026 papers fit this pattern: a first-author study and a middle-author paper, both from a small team at Tao's institution. The network is narrow but stable, typical of an early-stage researcher inside a single lab.
Activity77/100
Activity scores 77. OpenAlex countsByYear shows 6 works in 2024, a dip to 2 in 2025, and 5 in 2026, totaling 13 works in the recent three-year window. Both Evolve-reviewed papers are 2026 arXiv preprints, one with Tao as first author, so current output is concentrated in industrial robot perception. Despite the 2025 dip, the trajectory shows sustained recent publishing.
Reproducibility55/100
Reproducibility averages 55 across the 2 Evolve-reviewed papers, both rated medium readiness. Evolve notes the CAD-free 3D Gaussian Splatting shape-prior study has solid internal evidence but covers only 56 objects with no external baselines, while CALIPER is solid on its 18-part benchmark with broader generality unproven. Methods are described carefully, but independent external validation is not yet available.
Research summary
Chenxi Tao is an early-stage researcher at Georgia Institute of Technology working on robot perception for industrial automation, with a record built around one applied question: how to recognize visually similar industrial parts without CAD models. OpenAlex lists 13 works in the last three years (6 in 2024, 2 in 2025, 5 in 2026), an h-index of 2 and an i10-index of 0, marking a career at its beginning. Topic clusters center on Robot Manipulation and Learning (4 works) alongside advanced neural network applications, model reduction and neural networks, domain adaptation and few-shot learning, and industrial vision systems and defect detection. Two 2026 arXiv preprints have been reviewed by Evolve. In "Where Appearance Fails, Geometry Recognizes" Tao is first author, proposing a scan-derived 3D Gaussian Splatting shape prior that, combined with frozen vision foundation features, recovers CAD's recognition value without CAD onboarding — though evidence covers 56 objects with no external baselines. In CALIPER, a middle-author contribution, depth-derived metric size evidence complements DINOv2 support matching to separate near-identical parts on an 18-part benchmark. Both papers rate medium reproducibility (mean 55). Collaboration is narrow — 3 distinct co-authors, most often Alankrit Gupta and Seung-Kyum Choi — consistent with work inside a single lab.
Collaboration pattern
Collaboration is concentrated: 3 distinct co-authors across 13 works in the last three years, with Alankrit Gupta and Seung-Kyum Choi recurring most often. Both reviewed 2026 papers come from this small Georgia Tech group — Tao first author on the 3DGS shape-prior study and middle author on CALIPER. The low collaboration score (28) reflects a narrow but stable single-lab network rather than cross-institutional work.
Research keywords
Highlights
CALIPER: metric-grounded, model-free recognition of visually similar industrial parts
First-author study: scan-derived 3D Gaussian Splatting shape prior replaces CAD for recognition
A consistent agenda: removing CAD dependence from industrial object onboarding
Across both 2026 preprints, one question recurs — how to recognize visually near-identical industrial parts using only RGB-D input, learned geometry, and frozen foundation features, without CAD models at onboarding time.
Timeline
- 20246 works appear in the OpenAlex countsByYear record, marking the earliest year in the recorded publication window.
- 2025Recorded output dips to 2 works, the lowest annual count in the three-year window.
- 20265 recorded works; both Evolve-reviewed arXiv preprints appear — the CAD-free 3DGS shape-prior study (first author) and CALIPER (middle author).
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.
- Bibliometrics are drawn from OpenAlex only (identity: openalex); no Google Scholar or Scopus data is included.
- Only 2 of the 13 recent works have been reviewed by Evolve; quality verdicts cover just those two 2026 preprints.
- Output and impact dimensions are unscored because citation data is minimal (h-index 2, i10-index 0), consistent with a very early career stage.
- Chenxi Tao is a common name; this profile is name-resolved from public bibliographic data (OpenAlex/ORCID 0009-0007-4197-4062) and may not capture all works.