Table 2.

Comparison of segmentation performance for different methods.

Model	Detection Results				Training Time	F1/%
	mIoU	mPA	Precision	Recall
CA-DeepLabv3+	0.907	0.952	0.9488	0.9520	44h	95.1
DeepLabv3+	0.886	0.943	0.9352	0.9427	30h	93.9
UNet	0.722	0.777	0.8430	0.7775	26h	80.9
PSPNet	0.668	0.735	0.7703	0.7338	15h	71.5