Table 5. Comparison of deep learning algorithms.
| Model | FPS (GPU) | AP |
| RetinaNet | 22.4 | 0.630 |
| Faster-rcnn | 20.6 | 0.452 |
| YOLOv5-s | 60.9 | 0.926 |
| YOLOv5-m | 45.0 | 0.924 |
| YOLOv5-l | 41.1 | 0.922 |
| YOLOv5-x | 28.2 | 0.923 |
| Our Work | 40.6 | 0.981 |
| Model | FPS (GPU) | AP |
| RetinaNet | 22.4 | 0.630 |
| Faster-rcnn | 20.6 | 0.452 |
| YOLOv5-s | 60.9 | 0.926 |
| YOLOv5-m | 45.0 | 0.924 |
| YOLOv5-l | 41.1 | 0.922 |
| YOLOv5-x | 28.2 | 0.923 |
| Our Work | 40.6 | 0.981 |