Table 5 Comparison table of model training weight and training time.
From: Real-time detection of particleboard surface defects based on improved YOLOV5 target detection
Model | Model weight size (MB) | Training time (hours) | Detection of single image time (s) | Model parameter size | mAP @ 0.5 (%) | mAP @ 0.5: 0.95 (%) |
|---|---|---|---|---|---|---|
PB-YOLOv5s | 5.4 | 7.801 | 0.031 | 2.55 × 106 | 94.4 | 58.4 |
YOLOv5s | 14.4 | 9.613 | 0.047 | 7.07 × 106 | 92.1 | 53.2 |
PB-YOLOv5m | 14.0 | 15.221 | 0.043 | 6.78 × 106 | 96.3 | 67.8 |
YOLOv5m | 42.5 | 20.077 | 0.055 | 2.11 × 107 | 95.8 | 63.0 |
PB-YOLOv5l | 28.4 | 25.028 | 0.051 | 1.39 × 107 | 96.6 | 73.1 |
YOLOv5l | 93.8 | 54.328 | 0.060 | 4.67 × 107 | 97.8 | 69.3 |
PB-YOLOv5x | 49.7 | 57.583 | 0.057 | 2.46 × 107 | 98.3 | 76.7 |
YOLOv5x | 167.0 | 144.246 | 0.069 | 8.73 × 107 | 98.0 | 74.1 |
