Computer-vision researcher probing visible-infrared object detection from both attack and label-efficiency angles.
2
papers here
1
reviewed by Evolve
0
first-author
0
corresponding
—
works · OpenAlex
—
h-index · OpenAlex
Bibliometric figures come from OpenAlex (2026-09-21) 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 snapshot contains no publication-count or venue-volume data for output. The axis would measure total scholarly production over time — paper counts, venue tiers, and authorship roles — none of which are available here.
Impact—
Not yet scored: no citation counts, h-index, or other influence signals are present in this snapshot. The axis would measure how widely the work is cited and taken up by other researchers, which cannot be estimated here.
Focus100/100
Scored 100 for focus: every field-classified paper in this snapshot — one of one — falls under computer vision. The two reviewed 2026 works stay tightly on-theme, both targeting visible-infrared object detection, one through adversarial attack and one through semi-supervised learning. Within the reviewed sample this is exceptional single-field concentration, though the sample is small and broader output may diversify.
Collaboration59/100
Scored 59 for collaboration: the snapshot records 18 distinct coauthors, indicating steady team-based work rather than solo or isolated research. The frequent-coauthor list is led by Qi Ming, who appears twice, alongside Jiahuan Zhou, Juan Fang, Xiaxin Yuan, Xudong Zhao, Zhanchao Huang, Shaoguang Huang, Aleksandra Pizurica, Yueqi Zhu, Guo Cheng, and Yongkang Zhang. Both reviewed papers place Li in a middle author position, not corresponding author.
Activity—
Not yet scored: no year-by-year publication volume or longer temporal record exists beyond the two 2026 papers. The axis would measure publishing cadence and sustained research activity across years, which cannot be assessed from this snapshot.
Reproducibility55/100
Scored 55 for reproducibility, averaged over one assessed paper. Both reviewed works carry medium reproducibility readiness: the attack paper is limited by missing physical-world validation, while the semi-supervised paper reports unmatched baselines and single-run peak results. With a single-paper basis, this score should be read as an early signal rather than a settled track record.
Research summary
Jiangmeng Li is a computer-vision researcher whose reviewed work centers on visible-infrared (RGB-IR) object detection, approached from two complementary directions: exposing its vulnerabilities and reducing its annotation cost. Both cataloged papers appeared on arXiv in 2026. The first introduces CamoShift, an object-level attack that camouflages an adversarial patch and digitally shifts infrared evidence to misalign RGB-IR detectors; Evolve's review calls the results strong and broad while noting the absence of physical-world validation. The second is described as the first semi-supervised framework for image-pair-level RGB-IR detection, applying mean-teacher self-training and pseudo labels to oriented detection on weakly-aligned imagery, with strong results tempered by unmatched baselines and single-run peaks. Bibliometric indicators in this snapshot show a focus score of 100, with all field-classified output in computer vision; a collaboration score of 59 based on 18 distinct coauthors; and a reproducibility score of 55 averaged over a single assessed paper, with both works rated medium on reproducibility readiness. Li appears in a middle author position on both papers and is not listed as corresponding author, suggesting a contributor role within larger teams that include repeat collaborators such as Qi Ming, Jiahuan Zhou, and Aleksandra Pizurica. Output, impact, and activity metrics are not yet available for this profile.
Collaboration pattern
The snapshot lists 18 distinct coauthors and a collaboration score of 59. The most frequently recurring name is Qi Ming, appearing twice in the frequent-coauthor list, followed by Jiahuan Zhou, Juan Fang, Xiaxin Yuan, Xudong Zhao, Zhanchao Huang, Shaoguang Huang, Aleksandra Pizurica, Yueqi Zhu, Guo Cheng, and Yongkang Zhang. The mix suggests team-based computer-vision work spanning both attack and semi-supervised detection projects, with Li contributing in middle author positions rather than as corresponding author on the reviewed papers.
Research keywords
Highlights
CamoShift: object-level misalignment attacks on RGB-IR detectors
First semi-supervised framework for image-pair-level RGB-IR detection
A two-sided agenda: stress-testing and economizing the same modality pair
Taken together, the two 2026 papers treat visible-infrared detection as a single research object — first showing how cross-modal spatial misalignment can be weaponized, then asking how its annotation cost can be cut through semi-supervised teaching — with both works rated medium on reproducibility readiness.
Timeline
- 2026Published on arXiv the object-level misalignment attack (CamoShift) against visible-infrared object detectors, with strong and broad results pending physical-world validation.
- 2026Published on arXiv the aligned consensus teaching framework for label-efficient oriented object detection in weakly-aligned visible-infrared imagery.
- 2026Contributed to both the attack side and the learning side of RGB-IR detection within one year, as a middle author on both arXiv papers.
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.
- Dimension scores come from the platform's bibliometric snapshot; only focus (100), collaboration (59), and reproducibility (55) had sufficient underlying data, while output, impact, and activity are missing.
- No institution, country, or ORCID is attached to this record; the profile is name-resolved from public bibliographic data and may conflate researchers sharing the same name.
- The reproducibility score averages a single assessed paper, so it should be treated as an early-stage signal rather than a settled record.
- The narrative rests on only two reviewed arXiv papers (both 2026); venue acceptance, citation counts, and career facts fall outside this snapshot.