Table 4 Performance comparison of PEYOLO and YOLOv8 families.
From: PEYOLO a perception efficient network for multiscale surface defects detection
Methods | Params↓ | GFLOPS↓ | FPS↑ | \({\text{mAP}}_{50}\)↑ | F1 scores↑ | P↑ | R↑ |
|---|---|---|---|---|---|---|---|
YOLOv8n | 3.15 | 8.9 | 198.1 | 74.6 | 0.68 | 0.668 | 0.710 |
YOLOv8s | 11.16 | 28.8 | 170.4 | 74.2 | 0.69 | 0.698 | 0.689 |
YOLOv8m | 25.90 | 79.3 | 73.4 | 76.8 | 0.72 | 0.683 | 0.761 |
YOLOv8l | 43.69 | 165.7 | 42.7 | 77.8 | 0.71 | 0.749 | 0.698 |
YOLOv8x | 68.22 | 258.5 | 25.3 | 76.5 | 0.71 | 0.755 | 0.684 |
PEYOLO | 3.09 | 7.5 | 125.4 | 78.1 | 0.72 | 0.709 | 0.745 |
