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 1 不同算法在BIT交通数据集上的实验结果 |
| Tab.1 Experimental result of different algorithms on BIT traffic dataset |
|
| 模型 | AP0.5/% | mAP0.5/% | mAP0.5:0.95/% | | bus | microbus | minivan | sedan | suv | truck | | Yolov4[20] | 94.1 | 93.6 | 90.3 | 92.3 | 90.5 | 93.2 | 92.3 | 76.7 | | SSD[12] | 95.3 | 94.7 | 91.6 | 93.2 | 92.1 | 96.1 | 93.8 | 77.1 | | DERT[21] | 96.8 | 96.3 | 93.7 | 95.5 | 94.4 | 96.5 | 95.6 | 79.6 | | Faster RCNN[5] | 97.0 | 96.5 | 94.2 | 95.8 | 94.1 | 96.5 | 95.7 | 79.9 | | Cascade RCNN[7] | 97.8 | 97.2 | 94.3 | 96.7 | 94.9 | 98.1 | 96.5 | 80.4 | | Yolov5s[22] | 96.3 | 95.7 | 92.8 | 94.6 | 93.9 | 96.7 | 95.0 | 78.5 | | Yolov8s[23] | 98.3 | 97.8 | 95.1 | 97.5 | 95.7 | 98.2 | 97.1 | 81.2 | | SIAM-CML | 99.1 | 98.7 | 94.8 | 98.9 | 95.2 | 98.9 | 97.6 | 81.5 |
|
|
|