Table 4 Comparison between improved YOLOv5 and other mainstream object detection algorithms.

From: Automated crack detection of train rivets using fluorescent magnetic particle inspection and instance segmentation

Model

Precision (%)

Recall

mAP0.5 (%)

FPS

YOLOv5-x

89.1

79.8

86.8

13.2

Faster-RCNN32

81.8

73.5

79.0

3.7

YOLOX-x33

85.5

80.6

85.7

12.8

YOLOv6-l634

84.2

78.1

81.2

14.2

YOLOv7-x35

75.7

74.0

76.3

19.5

YOLOv8-x

88.2

84.1

89.7

20.6

YOLOv5x + ECSAM

 + NA + Soft-NMS

88.3

86.4

90.3

12.6

Back to article page