RAD-YOLO: a lightweight edge-deployed visual measurement method for small-object detection in UAV imagery

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Reliable visual measurement of small objects in unmanned aerial vehicle (UAV) imagery is challenging because targets often have low spatial resolution, dense spatial distribution, partial occlusion and strong background clutter. These factors reduce localisation accuracy and make real-time inference on embedded platforms difficult. To address the coupled requirements of measurement accuracy and deployability, this paper proposes RAD-YOLO, a YOLOv8s-based small-object detection method, and develops an edge-deployed variant named RAD-YOLO-Slim. An adaptive-field convolutional block attention module (AFCBAM) is first introduced into the backbone to enhance multi-scale receptive-field modelling and joint spatial-channel representation. A gated bidirectional asymptotic feature pyramid network (G-BiAFPN) is then designed to improve cross-scale semantic interaction while suppressing the propagation of shallow background responses. A distance-normalised Wise-IoU loss (D-WIoU) and soft non-maximum suppression (Soft-NMS) are further combined to refine bounding-box localisation and retain adjacent detections in dense small-object scenes. Finally, GSConv reconstruction, layer-adaptive magnitude-based pruning (LAMP) and 8-bit integer (INT8) post-training quantisation are used to construct RAD-YOLO-Slim for resource-constrained visual measurement on edge devices. Experiments on VisDrone2019 show that RAD-YOLO achieves 47.6% mean average precision at an IoU threshold of 0.5 (mAP@0.5), an improvement of 8.6 percentage points over YOLOv8s under the same protocol. RAD-YOLO-Slim maintains 44.1% mAP@0.5 with 5.42 M parameters and 6.2 giga floating-point operations (GFLOPs). After INT8 deployment on an RK3576 platform, RAD-YOLO-Slim achieves 43.1% mAP@0.5, 51 frames per second (FPS) and 11.81 FPS/W. These results indicate that the proposed method improves the reliability of object-level measurement outputs, including bounding-box localisation and countable detections, while maintaining practical inference speed and energy efficiency for embedded UAV vision systems.