Table 2 The experimental results of backbone network 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) |
|---|---|---|---|---|---|---|---|
YOLOv5n | 91.8 | 90.9 | 95.3 | 66.8 | 80.90 | 6.72 | 4.1 |
InceptionNeXtBlock | 89.9 | 89.8 | 94.2 | 63.2 | 75.68 | 6.71 | 4.6 |
FasterNext | 90.7 | 91.5 | 94.8 | 64.7 | 77.21 | 6.09 | 3.8 |
ShuffleNetV2Block | 79.0 | 75.6 | 82.5 | 45.2 | 81.77 | 3.09 | 1.5 |
BiFormerBlock | 90.5 | 91.4 | 95.1 | 67.4 | 61.73 | 8.05 | 7.5 |
CB2D | 91.8 | 90.3 | 95.5 | 67.6 | 63.16 | 7.07 | 4.3 |
ELANB | 88.9 | 88.1 | 93.6 | 61.9 | 76.36 | 5.29 | 3.1 |
ConXBv2 | 90.1 | 90.7 | 94.8 | 64.8 | 73.58 | 6.80 | 4.2 |
FocalNext | 92.5 | 92.5 | 95.8 | 67.8 | 77.94 | 6.64 | 4.2 |
