Carnegie Mellon researcher bridging computer vision and robotics — dense 3D tracking and humanoid motion generation
2
papers here
2
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
24
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-10-13) 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—
No output score is available yet. This axis would measure total publication volume, venue tiers, and authorship positions across the full record; the snapshot provides yearly counts and an i10-index of 44, but the calibrated output score is missing.
Impact—
No impact score is available yet. This axis would measure citation accumulation, citation velocity, and downstream influence; the snapshot lists raw indicators (h-index 24, i10-index 44) but has no calibrated impact score, so none is reported.
Focus0/100
Focus scores 0/100, the lowest value on the radar. It matches the raw field distribution — computer vision at 0.5 and robotics (humanoid motion generation and control) at 0.5 — and five OpenAlex topics of similar weight: advanced neural network applications (41), robot manipulation and learning (38), advanced vision and imaging (32), reinforcement learning in robotics (30), and 3D shape modeling and analysis (26).
Collaboration51/100
Collaboration scores 51/100, mid-range on the radar. The snapshot records 12 distinct coauthors in the recent window, with recurring partners including Ayush Jain, Sreeharsha Paruchuri, Ishita Gupta, Adam W. Harley, and Jakob Engel. On the two reviewed 2026 papers she appears as a middle author on TrackEverything and last author on HIGenNTO, consistent with contributing within larger research teams rather than solo efforts.
Activity100/100
Activity scores 100/100, the maximum on the radar. The raw series shows 37 works in the last three years and a marked acceleration: annual counts of 16 in 2023, 17 in 2024, and 14 in 2025, with 6 already logged for 2026 — the most intense stretch since the record began at 1 work in 2010.
Reproducibility38/100
Reproducibility scores 38/100, the average over the 2 papers Evolve has reviewed (reproMean 38). TrackEverything carries medium reproducibility readiness, supported by concrete benchmarks such as TAPVid-3D in its keyword set; HIGenNTO is rated low readiness. Both verdicts note dependence on narrow or internal baseline sets, which limits independent verification of the headline gains.
Research summary
Katerina Fragkiadaki is a researcher at Carnegie Mellon University whose work spans computer vision and robotics, with an OpenAlex-recorded field distribution split evenly (0.5 / 0.5) between computer vision and humanoid motion generation and control. Her bibliometric record shows an h-index of 24 and an i10-index of 44, with topic mass concentrated in advanced neural network applications (41), robot manipulation and learning (38), advanced vision and imaging (32), reinforcement learning in robotics (30), and 3D shape modeling and analysis (26). Annual counts in the snapshot climb from single digits through the 2010s to a recent peak — 16 works in 2023, 17 in 2024, and 14 in 2025 — with 37 works over the last three years, marking a highly active period. Evolve has reviewed two 2026 arXiv papers: TrackEverything, a dense long-horizon 3D point-tracking system built on voxel de-duplication of world-coordinate scene representations (middle authorship), and HIGenNTO, a noise-space trajectory optimization method for scalable humanoid interaction generation (last authorship). Both sit squarely within her recurring themes of 3D perception and learned robot control, though Evolve's verdicts note that each paper's headline claim rests on a limited or narrow baseline comparison, with reproducibility readiness rated medium and low respectively.
Collaboration pattern
The OpenAlex snapshot lists 12 distinct recent coauthors. Recurring collaborators in the record include Ayush Jain, Sreeharsha Paruchuri, Ishita Gupta, Lalit Jayanti, Adam W. Harley, Fan Zhang, Tanner Schmidt, Jakob Engel, Kashu Yamazaki, Yuto Shibata, and Kotaro Amaya. On the two reviewed 2026 papers she holds middle authorship (TrackEverything) and last authorship (HIGenNTO), a pattern consistent with senior contribution inside larger vision and humanoid-robotics teams rather than solo-first-author work.
Research keywords
Highlights
Dense 3D point tracking across 1000+ frames via voxel de-duplication
Noise-space trajectory optimization for scalable humanoid interaction generation
Timeline
- 2010First entries appear in the OpenAlex record (1 work); output stays low through 2016, ranging from 1 to 5 works per year.
- 2019Output steps up: 11 works in 2019 and 10 in 2020, following 7 in 2017 and 6 in 2018 — the busiest stretch to that point.
- 2023Recent peak: 16 works in 2023 and 17 in 2024, the two highest annual counts in the record, with 14 more in 2025.
- 2026Two Evolve-reviewed arXiv papers appear — TrackEverything (dense long-horizon 3D tracking, middle authorship) and HIGenNTO (humanoid interaction generation, last authorship); 6 works are logged for the year so far.
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 indicators (h-index 24, i10-index 44, yearly counts, topics, coauthor counts) come from an OpenAlex snapshot; per-paper citation counts and venue lists are not provided.
- The source record's country field reads 'RW', which does not correspond to the stated institution (Carnegie Mellon University); the affiliation region appears unresolved in the data.
- Reproducibility (38) is averaged over only 2 reviewed papers; it may not represent reproducibility across the full publication record.
- No ORCID is linked; the profile is name-resolved from public bibliographic data (identity: openalex), and authorship disambiguation relies on that record.