Table 5 The experimental results of loss function improvement.
From: YOLOFM: an improved fire and smoke object detection algorithm based on YOLOv5n
Model | Precision (%) | Recall (%) | mAP50 (%) | mAP50-95 (%) | FPS (%) | Params (MB) | GFLOPs (G) |
|---|---|---|---|---|---|---|---|
CIoU | 91.8 | 90.9 | 95.3 | 66.8 | 80.90 | 6.72 | 4.1 |
XIoU | 91.7 | 90.7 | 95.1 | 66.8 | 84.84 | 6.72 | 4.1 |
WIoU | 92.1 | 90.5 | 95.7 | 66.8 | 85.92 | 6.72 | 4.1 |
SIoU | 91.5 | 90.5 | 95.6 | 67.0 | 81.38 | 6.72 | 4.1 |
EIoU | 91.2 | 90.6 | 95.1 | 65.8 | 77.59 | 6.72 | 4.1 |
GIoU | 90.7 | 91.1 | 95.1 | 66.7 | 83.27 | 6.72 | 4.1 |
\(\alpha\)-IoU | 90.2 | 87.1 | 93.4 | 65.1 | 81.38 | 6.72 | 4.1 |
EfficiCIoU-Loss | 90.3 | 88.2 | 91.7 | 66.4 | 81.02 | 6.72 | 4.1 |
Focal-EIoU | 90.5 | 91.5 | 94.8 | 65.9 | 84.27 | 6.72 | 4.1 |
Focal-GIoU | 92.0 | 92.4 | 95.6 | 68.2 | 82.12 | 6.72 | 4.1 |
Focal-DIoU | 91.7 | 91.6 | 95.4 | 67.2 | 81.84 | 6.72 | 4.1 |
Focal-SIoU | 92.7 | 91.3 | 95.7 | 68.6 | 82.29 | 6.72 | 4.1 |
