Table 1 Ablation study on our attention module
From: Gastric polyp detection module based on improved attentional feature fusion
Method | MAP@0.5 (%) | F1 | P (%) | R (%) |
|---|---|---|---|---|
Yolov4+our feature fusion model | 91.52 | 0.87 | 94.94 | 80.07 |
Yolov4+our feature fusion model with spatial attention model | 91.03 | 0.87 | 93.42 | 80.78 |
Yolov4+our feature fusion model with position attention model | 92.17 | 0.89 | 93.65 | 83.99 |
Yolov4+our feature fusion model with channel attention model | 91.05 | 0.87 | 93.44 | 81.14 |
Yolov4+our feature fusion model with first channel then position attention model | 90.98 | 0.88 | 94.67 | 82.17 |
Yolov4+our attentional feature fusion model | 92.33 | 0.89 | 95.92 | 83.63 |