Table 3 The EAE-DETR dataset is compared with various other excellent methods in PKU-Market-PCB.
From: An improved EAE-DETR model for defect detection of server motherboard
Methods | mAP50 (%) | mAP50:95(%) | Parameters | GFLOPs |
|---|---|---|---|---|
Faster R-CNN | 80.1 | 49.2 | 41.3 M | 206.0 |
SSD | 79.2 | 48.8 | 62.8 M | 26.2 |
YOLOv3s-Tiny | 84.2 | 52.8 | 12.7 M | 19.0 |
YOLOv4s-Tiny | 85.2 | 54.3 | 6.06 M | 16.5 |
YOLOv5s | 91.2 | 59.2 | 7.02 M | 15.8 |
YOLOv7 | 85.2 | 56.1 | 36.48 M | 103.2 |
YOLOv8s | 92.3 | 61.3 | 11.27 M | 28.8 |
D-Fine-S | 91.8 | 61.7 | 10.18 M | 24.85 |
RTDETR-R18 | 93.4 | 63.2 | 19.8 M | 53.1 |
Ours | 96.1 | 65.1 | 15.56 M | 50.2 |
