融合流体自动标注与轻量化YOLOv8n的泥石流实时检测方法

王平,徐安之,赵洪黎,魏小源,杨富龙

Real-time debris flow detection method combining fluid automatic annotation and lightweight YOLOv8n

Ping WANG,Anzhi XU,Hongli ZHAO,Xiaoyuan WEI,Fulong YANG

表 5 相同数据集下的不同模型对比实验结果

Tab.5 Comparative experimental results of different models on same dataset

模型 P/% R/% mAP@0.5/% Params/M FLOPs/G FPS/(帧·s−1) Faster R-CNN 73.5 82.8 84.05 108.2 163.8 28.28 SSD 83.2 81.2 86.5 91.6 135.2 45.65 YOLOv5n 74.1 83.3 80.2 2.5 7.1 233.51 YOLOv8n 71.2 80.7 82.0 3.1 8.2 223.63 YOLOv8m 71.3 80.6 84.2 25.8 79.1 92.64 YOLOv10n 73.3 82.3 82.5 2.3 6.7 214.32 RefineDet[17] 70.2 76.8 78.2 26 10.2 56.27 RT-DETR-R18[18] 78.3 81.4 83.8 19.8 56.9 113.64 YOLOv8-Mudslide 72.3 82.8 86.7 2.6 7.3 230.89