基于对比学习的可扩展交通图像自动标注方法

侯越,李前辉,袁鹏,张鑫,王甜甜,郝紫微

Scalable traffic image auto-annotation method based on contrastive learning

Yue HOU,Qianhui LI,Peng YUAN,Xin ZHANG,Tiantian WANG,Ziwei HAO

表 2 不同算法在UA-DETRAC数据集上的实验结果

Tab.2 Experimental result of different algorithms on UA-DETRAC dataset

模型 AP0.5/% mAP0.5/% mAP0.5:0.95/% car bus van others Yolov4 82.1 80.3 71.5 69.7 75.9 54.5 SSD 85.3 84.7 73.1 75.7 79.7 58.8 DERT 88.4 87.6 74.2 82.9 83.3 62.5 Faster RCNN 88.7 87.9 74.3 83.1 83.5 62.9 Cascade RCNN 90.3 98.7 75.1 83.3 84.6 63.7 Yolov5s 88.1 87.3 74.2 82.4 83.0 61.6 Yolov8s 93.4 92.6 75.5 87.3 87.2 67.5 SIAM-CML 92.5 91.7 74.8 86.6 86.4 66.9