DSPE-ViT: a lightweight vision transformer with dynamic sparse positional encoding for dense small object detection in UAV imagery
Frontiers in Neurorobotics · 16.06.2026 · 20
DOI: 10.3389/fnbot.2026.1849093 ↗Kurz gefasst
Fachartikel in der Zeitschrift Frontiers in Neurorobotics, online erschienen am 16.06.2026. Verfasst von Liya Cai und 1 weiteren Autor:innen. Beteiligte Einrichtung: School of Intelligent Management, Suzhou Industrial Park Institute of Services Outsourcing, Suzhou, China. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 43 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Integrated Circuit and Artificial Intelligence Applications Industry-Education Integration Innovation Entity, Outstanding Teaching Team Project of “Qinglan Project” in Jiangsu Universities, Distinguished Teacher Workshop for IC Technology in Vocational Education of Jiangsu Province.
Abstract (Originalsprache)
Background
Detecting densely distributed small objects in unmanned aerial vehicle (UAV) aerial imagery poses a persistent challenge in computer vision. Vision Transformers (ViTs), empowered by global self-attention, perform strongly in object detection, but their fixed absolute positional encoding (PE) adapts poorly to scenes where small targets cluster at high density, and redundant encoding dimensions introduce unnecessary computational overhead.
Methods
This paper presents DSPE-ViT, a lightweight ViT-based detection framework tailored for dense small object detection in UAV imagery. Its core DSPE module comprises two complementary components: a PE Redundancy Pruner that employs learnable soft-gating masks to adaptively suppress redundant PE dimensions, and a Local PE Enhancer that introduces density-aware adaptive-window relative positional encoding to strengthen local spatial perception in high-density regions. Beyond the DSPE module, a Small Object Feature Pyramid Network (SmallObjFPN) integrating SE channel attention with depthwise separable convolutions improves multi-scale feature representation, and the WIoU v3 loss is adopted to refine bounding-box regression for small targets.
Results
On the VisDrone2019-DET dataset, DSPE-ViT achieves 43.2% mAP@0.5 with only approximately 6.0 M parameters and 15.8 GFLOPs. Cross-domain evaluation on SeaDronesSee yields 30.1% mAP@0.5 under zero-shot transfer and 38.4% after fine-tuning.
Conclusion
The cross-domain results confirm the generalization capability of the proposed lightweight framework.
Schlagwörter
Stichwörter der Autor:innen
Autor:innen und Institutionen
- School of Intelligent Management, Suzhou Industrial Park Institute of Services Outsourcing, Suzhou, China.
Förderung
- Integrated Circuit and Artificial Intelligence Applications Industry-Education Integration Innovation Entity
- Outstanding Teaching Team Project of “Qinglan Project” in Jiangsu Universities
- Distinguished Teacher Workshop for IC Technology in Vocational Education of Jiangsu Province
- Key Research Project of Jiangsu Provincial Higher Education Teaching Reform
Zitieren
APA
Cai, L., & Li, S. (2026). DSPE-ViT: a lightweight vision transformer with dynamic sparse positional encoding for dense small object detection in UAV imagery. Frontiers in Neurorobotics, 20. https://doi.org/10.3389/fnbot.2026.1849093
Vancouver
Cai L, Li S. DSPE-ViT: a lightweight vision transformer with dynamic sparse positional encoding for dense small object detection in UAV imagery. Frontiers in Neurorobotics. 2026;20. doi:10.3389/fnbot.2026.1849093
BibTeX
@article{Cai2026,
author = {Cai, Liya and Li, Shuping},
title = {DSPE-ViT: a lightweight vision transformer with dynamic sparse positional encoding for dense small object detection in UAV imagery},
journal = {Frontiers in Neurorobotics},
year = {2026},
volume = {20},
doi = {10.3389/fnbot.2026.1849093}
}
EUVIMED