EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Adapting Vision Language Pedestrian Re Identification across Cities via Uncertainty Reweighting

Daniel Chen, Michael Thompson, Sarah Nguyen

Global Media and Social Sciences Research Journal · 15.02.2026 · 7 (1) · 26-32

DOI: 10.71465/gmssrj141 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Global Media and Social Sciences Research Journal, online erschienen am 15.02.2026. Verfasst von Daniel Chen und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Autonomous driving deployments face open-set conditions where many pedestrians do not belong to any known identity gallery. Motivated by uncertainty-aware CLIP-based modal modeling, this paper introduces an uncertainty-calibrated similarity scoring scheme that supports both identification and unknown rejection. The framework estimates uncertainty from vision–language embedding disagreement and uses temperature-scaled calibration to improve decision reliability. Evaluations are performed on open-set protocols with 180,000 query images and mixed galleries where 30%–60% of queries correspond to unseen identities. Compared with softmax-threshold and distance-based open-set baselines using TransReID and CLIP embeddings, the proposed approach increases AUROC for unknown detection by 6.5%–9.2% while maintaining closed-set mAP within a 1.0% margin. At a fixed false accept rate, false matches drop by 20%–27% in heavy occlusion subsets.

Zitieren

APA

Chen, D., Thompson, M., & Nguyen, S. (2026). Adapting Vision Language Pedestrian Re Identification across Cities via Uncertainty Reweighting. Global Media and Social Sciences Research Journal, 7(1), 26-32. https://doi.org/10.71465/gmssrj141

Vancouver

Chen D, Thompson M, Nguyen S. Adapting Vision Language Pedestrian Re Identification across Cities via Uncertainty Reweighting. Global Media and Social Sciences Research Journal. 2026;7(1):26-32. doi:10.71465/gmssrj141

BibTeX
@article{Chen2026,
  author = {Chen, Daniel and Thompson, Michael and Nguyen, Sarah},
  title = {Adapting Vision Language Pedestrian Re Identification across Cities via Uncertainty Reweighting},
  journal = {Global Media and Social Sciences Research Journal},
  year = {2026},
  volume = {7},
  number = {1},
  pages = {26--32},
  doi = {10.71465/gmssrj141}
}
Kostenlos starten →