Table 3 Comparative experimental results of incorporating different attention mechanisms into the same convolutional module.
From: LPAE-YOLOv8: lightweight aerial small object detection via LSE-Head and adaptive attention
Models | mAP(0.5)(%) | mAP(0.5:0.95)(%) | The number of model parameters (×106) | GFLOPs (×109) |
|---|---|---|---|---|
YOLOv8-C(AA) | 33.5% | 19.5% | 3.02 | 9.8 |
YOLOv8-C(CA) | 32.3% | 18.9% | 3.02 | 9.8 |
YOLOv8-C(SA) | 32.6% | 19.0% | 3.02 | 9.8 |
YOLOv8-C(GC) | 32.2% | 18.9% | 3.02 | 9.8 |
