Table 4 Verify the applicability of our attentional feature fusion module on the up-to-date backbone networks
From: Gastric polyp detection module based on improved attentional feature fusion
Method | MAP@0.5 (%) | P (%) | R (%) |
|---|---|---|---|
Shufflenetv2+FPN | 78.70 | 85.90 | 71.40 |
Shufflenetv2+PANet | 81.40 | 85.30 | 75.70 |
Shufflenetv2+our attentional feature fusion model | 83.10 | 86.40 | 77.10 |
Mobilenetv3+FPN | 76.50 | 79.80 | 69.00 |
Mobilenetv3+PANet | 80.50 | 80.60 | 77.10 |
Mobilenetv3+our attentional feature fusion model | 81.40 | 81.40 | 77.10 |
GhostBottleneck+FPN | 76.30 | 85.40 | 70.00 |
GhostBottleneck+PANet | 83.40 | 92.10 | 77.60 |
GhostBottleneck+our attentional feature fusion model | 85.10 | 92.60 | 77.60 |
Res2Net50+FPN | 79.50 | 87.20 | 68.60 |
Res2Net50+PANet | 83.30 | 77.10 | 77.60 |
Res2Net50+our attentional feature fusion model | 84.00 | 91.00 | 78.60 |