Table 2 The EAE-DETR dataset is compared with various other excellent methods in in PCBA-DET.
From: An improved EAE-DETR model for defect detection of server motherboard
Methods | mAP50 (%) | mAP50:95(%) | Precision(%) | Recall(%) | Parameters | GFLOPs |
|---|---|---|---|---|---|---|
YOLOv8s | 74.1 | 47.2 | 72.2 | 71.6 | 11.27 M | 28.8 |
Faster-RCNN27 | 70.1 | 23.5 | 64.2 | 63.1 | 41.3 M | 206 |
RTMDet-Tiny28 | 69.8 | 23.2 | 64.1 | 62.2 | 4.86 M | 8.13 |
RetinaNet29 | 64.2 | 21.5 | 58.8 | 58.3 | 36.6 M | 93.7 |
YOLOv5s | 71.2 | 25.4 | 69.5 | 68.1 | 7.3 M | 16.2 |
YOLOv10n | 70.1 | 25.6 | 68.8 | 68.1 | 2.23 M | 6.6 |
SSD30 | 50.2 | 17.2 | 40.6 | 38.2 | 26.21 M | 85.3 |
YOLOv6m | 68.3 | 24.7 | 62.7 | 60.6 | 34.5 M | 82.2 |
YOLOX-Tiny31 | 66.2 | 23.1 | 61.9 | 60.8 | 5.01 M | 7.57 |
YOLOv9s | 71.2 | 26.5 | 70.2 | 68.4 | 9.6 M | 25.3 |
MELF-YOLOv5s32 | 62.6 | 22.1 | 55.1 | 53.8 | 3.5 M | 8.1 |
YOLO11n | 70.8 | 25.3 | 68.3 | 66.4 | 2.62 | 6.4 |
YOLO11s | 74.7 | 28.7 | 72.3 | 71.4 | 9.43 M | 21.5 |
D-Fine-S | 71.2 | 25.1 | 69.1 | 67.4 | 10.18 M | 24.85 |
DEIM-S | 70.5 | 24.3 | 68.4 | 67.1 | 10.26 M | 24.68 |
RTDETRV2-R18 | 73.3 | 27.5 | 71.2 | 68.8 | 20 M | 60 |
PCBA-YOLO33 | 72.4 | 26.3 | 70.1 | 68.2 | 12.4 M | 20.8 |
YOLO-MBBi34 | 70.2 | 25.5 | 69.4 | 66.1 | 6.2 M | 12.8 |
Ours | 78.5 | 32.6 | 76.8 | 75.4 | 15.56 M | 50.2 |
