面向自动驾驶的轻量化交警手势识别方法
|
|
柳长源,赵海健,吴海滨,刘佳伟
|
Lightweight traffic police gesture recognition method for autonomous driving
|
|
Changyuan LIU,Haijian ZHAO,Haibin WU,Jiawei LIU
|
|
| 表 2 各主流目标检测算法性能对比实验结果 |
| Tab.2 Performance comparison experiments of various mainstream target detection algorithms |
|
| 网络模型 | GFLOPs/109 | Params/106 | FR/(帧∙s−1) | P/% | R/% | mAP/% | | RT-DETR-R18[13] | 10.1 | 18.7 | 180 | 92.8 | 91.4 | 90.2 | | Gold-YOLO-N[14] | 12.1 | 5.6 | 265 | 96.0 | 95.1 | 95.2 | | YOLOv5s[15] | 16.5 | 7.2 | 233 | 93.0 | 92.7 | 94.2 | | YOLOv8s[16] | 28.6 | 11.2 | 206 | 97.5 | 96.0 | 96.0 | | YOLOv9s[17] | 26.7 | 7.2 | 243 | 96.3 | 93.5 | 95.1 | | YOLOv10s[18] | 21.6 | 7.2 | 264 | 96.8 | 97.3 | 96.6 | | YOLOv11s[19] | 21.5 | 9.4 | 258 | 97.8 | 98.4 | 97.6 | | 本研究网络 | 5.7 | 1.8 | 294 | 99.2 | 98.6 | 97.9 |
|
|
|