Table 7 Performance difference between pretrained and non-pretrained models
Model (Large CSPDarknet as backbone) | mAP at 0.5 % | mAP at 0.5:0.9 % |
|---|---|---|
Improved framework with pretrained | 97.07 (P=0.63) | 64.32 (P=0.0017) |
Original YOLOv5 with pretrained | 96.10 | 58.58 |
Improved framework without pretrained | 96.82 | 59.73 |
Original YOLOv5 without pretrained | 96.18 | 58.15 |