A New Steel Defect Detection Algorithm Based on Deep Learning

Table 6

The algorithms of SSD, RetinaNet, and Cascade R-CNN are compared with the experimental results of the proposed algorithm.

Algorithms	AP						mAP	mAP increasing
	a	b	c	d	e	f
SSD	0.302	0.522	0.620	0.390	0.709	0.515	0.510	+0.242
RetinaNet	0.391	0.676	0.772	0.711	0.546	0.542	0.606	+0146
Cascade R-CNN	0.321	0601	0.794	0.723	0.509	0.805	0.626	+0.126
This paper	0.501	0.791	0.792	0.874	0.649	0.905	0.752	−