Table 3.
Detection results of different algorithms on VisDrone data set.
| Algorithm Model | Awn-tr | Bicycle | Bus | Car | Motor | Pedestrian | People | Truck | Tricycle | Van | mAP0.5 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Fast R-CNN | 8.73 | 5.8 | 43.8 | 44.1 | 16.8 | 12.5 | 8.1 | 30.4 | 8.5 | 20.4 | 19.9 |
| CenterNet [35] | 14.2 | 7.5 | 42.6 | 61.9 | 18.8 | 22.9 | 11.6 | 24.7 | 13.1 | 19.4 | 23.7 |
| SSD | 11.1 | 7.3 | 49.8 | 63.2 | 19.1 | 18.7 | 9.0 | 33.1 | 11.7 | 29.9 | 25.3 |
| DMNet [36] | 14.1 | 8.9 | 49.2 | 58.9 | 29.3 | 27.6 | 18.9 | 29.3 | 20.3 | 30.2 | 28.7 |
| YOLOv3 | 7.7 | 6.8 | 39.3 | 68.8 | 21.5 | 22.5 | 12.5 | 26.4 | 8.4 | 24.3 | 23.8 |
| YOLOv4 | 12.3 | 8.6 | 48.8 | 69.2 | 22.7 | 26.6 | 14.5 | 29.9 | 12.6 | 27.2 | 27.3 |
| YOLOv6 | 10.2 | 5.0 | 43.0 | 74.1 | 32.4 | 31.4 | 25.5 | 26.7 | 18.1 | 35.7 | 30.2 |
| Ours | 15.1 | 11.6 | 51.1 | 80.9 | 45.4 | 43.8 | 34.5 | 30.8 | 25.2 | 44.2 | 38.3 |