Fig 1.
Structure of YOLOv11.
Fig 2.
Structure of DMS-YOLO.
Fig 3.
Structure of C3k2-DMS-EdgeNet.
Fig 4.
Structure of DMS-EdgeNet.
Fig 5.
Structure of C2f-DMS-EdgeNet.
Fig 6.
Structure of DySAN.
Fig 7.
Dynamic scaling factor generation via SE attention.
Fig 8.
Target size distribution of the dataset.
(a): Distribution of target sizes for the dataset Aerial Traffic Images. (b): Distribution of target sizes for the dataset VisDrone-DET2019.
Table 1.
Experimental environment.
Table 2.
Experimental hyperparameters.
Table 3.
Results of the position experiment.
Table 4.
Results of ablation experiments.
Table 5.
Results of comparative experimental.
Table 6.
Comparison of detection results of different algorithms on the dataset VisDrone-DET2019.
Fig 9.
Comparison of mAP50-small across different YOLO versions.
Fig 10.
Precision-Recall curve comparison.
(a): DMS-YOLO (b): YOLOv11n.
Fig 11.
Detection results in different scenarios.
(a): DMS-YOLO detection results. (b): YOLOv11n detection results.
Fig 12.
(a): DMS-YOLO detection results. (b): YOLOv11n detection results.