Moaz Amjad

National University of Sciences and Technology

Unclaimed · AI-structured profileOpenAlex

NUST researcher in dental-radiograph segmentation and integer-only test-time adaptation for microcontrollers

papers here

reviewed by Evolve

first-author

corresponding

—

works · OpenAlex

—

h-index · OpenAlex

Bibliometric figures come from OpenAlex (2026-10-17) 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—

This axis is currently unscored: the bibliometric snapshot provides no publication counts or productivity metrics for Moaz Amjad. Output would measure cumulative and per-year scholarly production — preprints, peer-reviewed papers, and datasets — over the researcher's career.

Impact—

Impact is not scored: no citation counts, h-index, or citation-derived indicators exist in the current data. This axis would measure how frequently the community cites and builds upon this researcher's publications across venues and time.

Focus100/100

Focus scores 100. The underlying field distribution maps the indexed record to a single field — medical image analysis (computer vision, dental radiography), with a count of 1 — so measured concentration is maximal, though it rests on a very small sample. The second catalog paper extends into TinyML and efficient edge inference, suggesting breadth beyond what the single-field statistic captures.

Collaboration41/100

Collaboration scores 41, derived from 7 distinct co-authors in the bibliometric snapshot — a compact network. Muhammad Rehan and Haider Ali each recur in the co-author list, alongside Syed Danial Ahmed, Mariam Adnan, and Muhammad Ali Munir. Across the two reviewed papers Amjad appears in middle and last author positions and is corresponding author on neither, consistent with contributing inside a small, stable team.

Activity—

Activity is not scored: the snapshot lacks year-by-year publication timestamps, so recent publishing cadence cannot be computed. It would measure output momentum across consecutive years, currently evidenced only by two 2026 arXiv preprints.

Reproducibility55/100

Reproducibility scores 55, the mean readiness across 1 scored paper (reproPapers = 1, reproMean = 55). Both reviewed preprints carry a medium repro-readiness rating: the panoramic-radiograph segmentation study runs rigorous ablations but a private corpus blocks independent verification, while FORGE offers strong hardware validation yet a small-scale evaluation and a safety gate failing on many classes.

Research summary

Moaz Amjad is a researcher at the National University of Sciences and Technology (NUST) whose reviewed work spans two applied machine-learning directions: medical image analysis of dental radiography and efficient inference on microcontroller hardware. Evolve's catalog contains two 2026 arXiv preprints. The first, a middle-author study of tooth segmentation on panoramic radiographs, argues that detection is effectively solved and that image resolution rather than architecture governs delineation quality; Evolve's review credits rigorous ablations while noting that a private corpus blocks independent verification. The second, FORGE, where Amjad is last author, proposes forward-only test-time adaptation for integer-only vision models on microcontrollers; the review highlights strong hardware validation but flags a small-scale evaluation and a safety gate that fails on many classes. Bibliometric indicators are narrow: focus scores 100 on a single indexed field, collaboration scores 41 across 7 distinct co-authors, and reproducibility averages 55 from one scored paper, while output, impact, and activity lack sufficient data and remain unscored. Recurring co-authors include Muhammad Rehan and Haider Ali, alongside Syed Danial Ahmed, Mariam Adnan, and Muhammad Ali Munir. This is an early-stage profile built from a limited snapshot rather than a complete publication record.

Collaboration pattern

Collaboration scores 41 from 7 distinct co-authors, indicating a compact network. Muhammad Rehan and Haider Ali recur in the co-author list, joined by Syed Danial Ahmed, Mariam Adnan, and Muhammad Ali Munir. On the two catalog papers Amjad holds middle and last authorship respectively, neither as corresponding author, consistent with contributing within a small NUST-affiliated team across dental imaging and edge-inference projects.

Research keywords

panoramic radiographytooth segmentationFDI numberinginstance segmentationdomain shiftSAM2test-time adaptationmicrocontrollersTinyMLquantizationinteger-only inferencebatch normalization folding

Highlights

Resolution, not architecture, governs tooth segmentation on panoramic radiographs
In this middle-author 2026 arXiv study, Evolve's review finds the central claim — that detection is solved while delineation is not — is supported by rigorous ablations; a private corpus, however, prevents independent verification (repro readiness: medium).
FORGE revives test-time adaptation on int8 microcontroller vision models
As last author, Amjad contributed to this 2026 arXiv proposal of forward-only recalibration that restores test-time adaptation on batch-normalization-folded integer-only models; Evolve credits strong hardware validation while flagging a small-scale evaluation and a safety gate that fails on many classes.
Medium repro-readiness across both reviewed preprints
Both Evolve-reviewed papers are rated medium for reproducibility readiness, producing the reported mean of 55 over one scored paper; independent replication of either line of work currently hinges on future code or data release.

Timeline

2026Middle-author arXiv preprint on panoramic-radiograph tooth segmentation argues that resolution, not architecture, governs delineation quality.
2026Last-author arXiv preprint FORGE proposes forward-only test-time adaptation for integer-only vision models on microcontrollers.
2026Both catalog preprints appear in the same year, spanning medical image analysis and TinyML / efficient edge inference.

Papers in the Evolve catalog

Frequent co-authors

Muhammad Rehan×1 Syed Danial Ahmed×1 Mariam Adnan×1 Haider Ali×1 Muhammad Rehan×1 Haider Ali×1 Muhammad Ali Munir×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 dimension scores come from a limited OpenAlex-based bibliometric snapshot; the output, impact, and activity axes lack underlying data and are not scored.
Only two Evolve-reviewed arXiv preprints (both 2026) inform this profile; it does not represent a complete publication record.
No ORCID or country is available, so identity is resolved by name and institutional affiliation; Moaz Amjad may not be a unique name.
The reproducibility mean of 55 rests on a single scored paper, and the focus score of 100 rests on one indexed field assignment, so both are sample-sensitive.