Class-Conditioned Evidence Recovery in Frozen Vision Transformers for Remote Sensing Object Detection
Remote Sensing · 16.11.2026 · 18 (22) · 3929
DOI: 10.3390/rs18223929 ↗Kurz gefasst
Fachartikel in der Zeitschrift Remote Sensing, online erschienen am 16.11.2026. Verfasst von Haoxuan Xiang und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Aerial object detectors must resolve small, closely spaced objects while retaining enough context to recognize large structures. RS-Adapter uses an image-level class prior to update intermediate features in a frozen vision transformer. The prior combines a learned classifier, fixed class prototypes, and class-conditional descriptor statistics. Cross-attention links the prior to local features, which are fused across scales and gated before being added to the transformer. The adapted features guide a denoiser that generates box prompts for the detection decoder. On VisDrone, RS-Adapter achieved 39.2% average precision (AP) and 28.4% small-object AP, 3.1 percentage points above DEIM-D-FINE-L on the latter metric. AP50 was 78.3% on DIOR, 84.5% on NWPU VHR-10, and 87.7% on RSOD. RS-Adapter trained 16.8 million of its 103.2 million parameters and used less peak training memory than full tuning in the VisDrone comparison.
Zitieren
APA
Xiang, H., Zhao, X., Yang, S., & Lin, P. (2026). Class-Conditioned Evidence Recovery in Frozen Vision Transformers for Remote Sensing Object Detection. Remote Sensing, 18(22), 3929. https://doi.org/10.3390/rs18223929
Vancouver
Xiang H, Zhao X, Yang S, Lin P. Class-Conditioned Evidence Recovery in Frozen Vision Transformers for Remote Sensing Object Detection. Remote Sensing. 2026;18(22):3929. doi:10.3390/rs18223929
BibTeX
@article{Xiang2026,
author = {Xiang, Haoxuan and Zhao, Xiaodong and Yang, Sheng and Lin, Peng},
title = {Class-Conditioned Evidence Recovery in Frozen Vision Transformers for Remote Sensing Object Detection},
journal = {Remote Sensing},
year = {2026},
volume = {18},
number = {22},
pages = {3929},
doi = {10.3390/rs18223929}
}
EUVIMED