Frozen DINOv2–CLIP Fusion and Binary Hashing for Remote Sensing Image Retrieval: A Controlled Evaluation
Journal of Computing Theories and Applications · 13.08.2026 · 4 (1) · 354-368
DOI: 10.62411/jcta.16661 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Computing Theories and Applications, online erschienen am 13.08.2026. Verfasst von Jumi J und 3 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
This study evaluates a controlled content-based remote sensing image retrieval pipeline that combines frozen vision foundation models with binary hashing. Across five stratified seeds on PatternNet and NWPU-RESISC45, calibrated DINOv2–CLIP score fusion achieved mAP values of 0.8143 ± 0.0007 and 0.5693 ± 0.0020, respectively, improving over the stronger single backbone by 0.0582 and 0.0514. The two similarity spaces were related but not redundant, with Spearman correlations of ρ = 0.510 and 0.515 across 200,000 sampled pairs. The contribution of texture features was dataset- and backbone-dependent: DINOv2–texture received a weight of 0.1 on PatternNet, whereas CLIP–texture and three-stream fusion assigned zero weight to texture; forcing texture into a 32-bit representation reduced mAP by 0.0147 and 0.0510. Among the unsupervised hashing methods, 64-bit iterative quantization (ITQ) exceeded the full-precision fused baseline by 0.0533 and 0.0445 mAP on PatternNet and NWPU-RESISC45, respectively. The corresponding paired parametric tests were significant, although the exact five-pair sign-flip test remained resolution-limited (p = 0.0625). A supervised CSQ-style head achieved mAP values of 0.9886 and 0.9087 at 64 bits; however, this improvement cannot be attributed to compression alone because gallery labels were used during training. In a controlled 200,000-vector benchmark constructed through deterministic repetition, 64-bit flat Hamming search required 1.6 MB and 0.674 ms/query, compared with 1.02 GB and 5.343 ms/query for 1,280-dimensional float vectors. Overall, the results support DINOv2–CLIP fusion and 64–128-bit ITQ as practical label-free choices, while restricting the scalability claim to the tested flat-search setting.
Zitieren
APA
J, J., Mulyono, T., Zaenuddin, A., & Suwardi, S. (2026). Frozen DINOv2–CLIP Fusion and Binary Hashing for Remote Sensing Image Retrieval: A Controlled Evaluation. Journal of Computing Theories and Applications, 4(1), 354-368. https://doi.org/10.62411/jcta.16661
Vancouver
J J, Mulyono T, Zaenuddin A, Suwardi S. Frozen DINOv2–CLIP Fusion and Binary Hashing for Remote Sensing Image Retrieval: A Controlled Evaluation. Journal of Computing Theories and Applications. 2026;4(1):354-368. doi:10.62411/jcta.16661
BibTeX
@article{J2026,
author = {J, Jumi and Mulyono, Tedjo and Zaenuddin, Achmad and Suwardi, Suwardi},
title = {Frozen DINOv2–CLIP Fusion and Binary Hashing for Remote Sensing Image Retrieval: A Controlled Evaluation},
journal = {Journal of Computing Theories and Applications},
year = {2026},
volume = {4},
number = {1},
pages = {354--368},
doi = {10.62411/jcta.16661}
}
EUVIMED