Table 1 Performance comparison of different object detection models.

From: Confidence–gradient reweighting and lightweight feature enhancement algorithm for steel surface defect detection

Model

mAP@0.5:

0.95 (%)

mAP@0.5

(%)

P

(%)

R

(%)

F1

(%)

Param

(M)

Latency

(ms/img)

YOLOv5s

42.51

73.24

73.15

69.82

71.45

7.20

14.11

YOLOv8s

39.97

72.14

74.57

66.98

70.57

11.16

12.52

YOLO11s

42.66

74.69

70.72

73.66

72.16

9.44

15.97

RT-DETR-L

37.52

68.97

69.42

64.14

66.68

32.00

43.47

Faster R-CNN (R50-FPN)

33.82

66.25

59.49

71.19

64.82

43.71

30.10

(GRACE) Ours

43.66

75.88

73.62

72.33

72.97

9.56

16.93

Back to article page