基于改进RT-DETR的复杂天气下铁路异物实时检测算法
|
|
牛宏侠,冯鼎超,侯涛
|
Real-time detection algorithm for railway foreign objects in complex weather conditions based on improved RT-DETR
|
|
Hongxia NIU,Dingchao FENG,Tao HOU
|
|
| 表 5 不同目标检测算法对比实验结果 |
| Tab.5 Comparison of results from different object detection algorithms |
|
| 方法 | mAP@0.5/% | mAP@0.5:0.95/% | Parameters | FLOPs/109 | FPS/(帧·s−1) | | RT-DETR-r18 | 94.02(±0.3) | 76.05(±0.3) | 19880748 | 57.0 | 75 | | SSD | 81.13(±0.12) | 64.23(±0.15) | 2657893 | 2.6 | 55 | | Faster R-CNN | 90.17(±0.11) | 69.58(±0.14) | 28304605 | 908.9 | 14 | | Sparse R-CNN | 89.80(±0.08) | 67.46(±0.06) | 77834677 | 23.2 | 48 | | YOLOv8m | 92.44 | 74.58 | 25908746 | 79.3 | 101 | | YOLOv8l | 92.87(±0.1) | 74.84(±0.1) | 43691578 | 168.0 | 68 | | YOLOv11m | 92.11(±0.1) | 76.01(±0.1) | 20001579 | 67.7 | 94 | | YOLOv11l | 92.99(±0.1) | 75.03(±0.1) | 25414597 | 87.0 | 64 | | TOOD[24] | 88.08(±0.1) | 70.64(±0.1) | 32001587 | 199.0 | 23 | | DEIM[25] | 93.58(±0.2) | 76.06(±0.2) | 19264887 | 56.7 | 81 | | 文献[26] | 92.88 | 74.58 | 19847549 | 63.1 | 121 | | 本研究算法 | 95.67(±0.19) | 79.45(±0.3) | 7264632 | 56.8 | 88 |
|
|
|