Figure 1.
YOLOv10 model structure
Figure 2.
APFU-YOLOv10 model structure
Figure 3.
APFBmodel structure
Figure 4.
APU model structure
Figure 5.
Structure of FEA module.
TABLE I.
COMPARISON BEFORE AND AFTER ALGORITHM IMPROVEMENT
| Class | YOLOv10 | APFU-YOLOv10 |
|---|---|---|
| Layers | 237 | 253 |
| Parameters | 7.22M | 7.70M |
| GFLOPs | 21.4 | 21.8 |
| Box Precision (P) | 0.530 | 0.539 |
| Recall (R) | 0.405 | 0.417 |
| mAP@0.5 | 0.426 | 0.435 |
| mAP@0.5:0.95 | 0.258 | 0.262 |
| Inference Speed | 7.8ms | 8.2ms |
Figure 6.
Comparison chart of mAP during training
Figure 7.
Average precision of each label in YOLOv10
Figure 8.
Average precision of each label in APFU-YOLOv10