基于YOLOv8s的轻量化航拍图像小目标检测算法

邬开俊,郑云琦,魏鼎,袁海翔

YOLOv8s based lightweight algorithm for small object detection in aerial imagery

Kaijun WU,Yunqi ZHENG,Ding WEI,Haixiang YUAN

表 2 YOLOv8其他模型改进后在VisDrone验证集上的检测性能对比结果

Tab.2 Comparison results of detection performance of other improved YOLOv8 models on VisDrone validationset

方法 AP/% AP50/% AP75/% APS/% APM/% APL/% FLOPs/109 Params/106 YOLOv8n 17.9 31.1 17.9 9.6 27.6 35.7 6.8 2.68 YOLOv8n+DMSA-Net+TDFM+P2+ECSAM 21.3 35.8 21.8 12.4 30.9 37.1 12.2 2.51 YOLOv8m 24.7 41.0 25.2 15.3 36.8 43.5 79.1 25.85 YOLOv8m+DMSA-Net+TDFM+P2+ECSAM 28.4 46.9 29.0 19.1 40.1 45.3 101.8 21.89 YOLOv8l 26.5 43.3 27.3 16.2 39.4 49.4 164.9 43.6 YOLOv8l+DMSA-Net+TDFM+P2+ECSAM 30.1 49.4 30.8 21.0 42.0 50.1 213.0 39.2 YOLOv8x 27.4 44.4 28.4 17.0 41.1 47.5 257.4 68.13 YOLOv8x+DMSA-Net+TDFM+P2+ECSAM 30.7 50.1 31.5 21.8 42.6 47.8 328.7 61.05