FRFDet: Efficient UAV Small Object Detection with Symmetric Sampling and Scalable Fusion

2026-07-05

AI Summary

This work addresses the challenges of detecting small objects in drone-captured images under adverse conditions such as complex weather, low illumination, and sensor noise—issues often exacerbated by cluttered backgrounds, degraded details, and ineffective multi-scale feature fusion. To this end, the authors propose FRFDet, a lightweight single-stage detector featuring two key innovations: a plug-and-play Inverse Bidirectional Sampling (IBS) module for spatial detail reconstruction, and a Scale–Feature Relation Cross-Fusion (SFRCF) module that reveals an important relationship between model scale and fusion strategy—element-wise multiplication suits compact models, while addition-based fusion benefits larger ones. Integrating a channel expansion–compression architecture with a scale-aware cross-group fusion mechanism, FRFDet achieves state-of-the-art performance among lightweight detectors on VisDrone, UAVDT, HazyDet, and MS COCO, offering low computational cost, minimal parameters, and high inference speed suitable for resource-constrained UAV platforms.

Abstract

Small object detection in Unmanned Aerial Vehicle (UAV) imagery remains challenging under adverse conditions, including complex weather, low illumination, and sensor noise. These challenges mainly stem from severe background clutter, fine-grained detail degradation, and suboptimal semantic-spatial feature fusion, which jointly hinder robust small-object representation. To this end, we propose FRFDet, a lightweight yet effective single-stage detector tailored for UAV-based small object detection. FRFDet proposes two plug-and-play modules: Inverse Bidirectional Sampling (IBS) and Scale-Feature Relationship Cross-Fusion (SFRCF). IBS preserves critical spatial details via channel expansion-compression and bidirectional pattern reconstruction, improving feature alignment. SFRCF explicitly models scale-dependent fusion behaviors, revealing that inter-group element-wise multiplication favors compact models, while inter-group additive fusion benefits larger architectures. Extensive experiments on VisDrone, UAVDT, HazyDet, and MS COCO demonstrate that FRFDet achieves state-of-the-art performance among lightweight detectors with low computational cost, compact parameters, and fast inference, making it well suited for resource-constrained UAV platforms.

Problem

• small object detection
• UAV imagery
• adverse conditions
• feature fusion
• background clutter

Innovation

• Inverse Bidirectional Sampling
• Scale-Feature Relationship Cross-Fusion
• lightweight object detection
• UAV small object detection
• feature fusion