Table 3 Comparison of detection results of different models on GC10-DET dataset.
From: Feature optimization-guided high-precision and real-time metal surface defect detection network
Methods | Backbone | mAP(%) | AP(%) | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
Pu | Wl | Cg | Ws | Os | Ss | In | Rp | Cr | Wf | |||
RetinaNet30 | ResNet50 | 59.9 | 92.4 | 88.4 | 94.5 | 74.1 | 54.5 | 54.4 | 28.7 | 15.5 | 21.4 | 75.1 |
Faster R-CNN32 | VGG16 | 41.3 | 66.1 | 26.4 | 91.3 | 56.6 | 49.5 | 43.5 | 8.9 | 6.1 | 14.3 | 47.6 |
Faster R-CNN32 | ResNet50 | 60.8 | 82.2 | 78.0 | 95.4 | 69.2 | 57.7 | 58.3 | 24.8 | 29.2 | 30.7 | 82.6 |
YOLOv3-spp55 | Darknet53 | 60.6 | 96.5 | 82.5 | 96.8 | 75.5 | 57.4 | 48.4 | 26.4 | 22.0 | 14.4 | 77.6 |
YOLOv3-tiny23 | Darknet19 | 59.7 | 96.3 | 65.6 | 97.8 | 80.1 | 60.6 | 45.1 | 22.0 | 21.6 | 41.0 | 67.2 |
YOLOv424 | CSPDarknet53 | 61.2 | 90.4 | 89.8 | 93.9 | 62.6 | 59.4 | 48.3 | 23.6 | 17.7 | 37.6 | 88.2 |
YOLOv5s | CSPDarknet53 | 64.2 | 96.0 | 87.9 | 97.0 | 77.4 | 60.5 | 56.5 | 21.6 | 28.6 | 36.5 | 79.6 |
YOLOv726 | E-ELAN | 65.3 | 95.8 | 74.1 | 94.1 | 81.9 | 57.2 | 57.5 | 25.4 | 40.6 | 47.1 | 79.4 |
YOLOX56 | CSPDarknet53 | 59.9 | 89.7 | 89.8 | 89.6 | 67.9 | 61.0 | 57.2 | 28.6 | 27.9 | 25.9 | 61.1 |
YOLOv8s | CSPDarknet53 | 66.2 | 98.3 | 92.3 | 97.1 | 76.8 | 65.9 | 49.9 | 24.1 | 28.6 | 46.7 | 82.8 |
RT-DETR57 | HGNetv2 | 64.4 | 97.8 | 90.2 | 95.8 | 72.6 | 57.8 | 45.3 | 26.8 | 37.5 | 45.4 | 74.4 |
FOHR Net | CSPDarknet53 | 70.5 | 95.9 | 91.4 | 96.2 | 82.7 | 63.1 | 61.4 | 29.6 | 55.4 | 48.5 | 80.9 |
