Semisupervised Semantic Segmentation with Mutual Correction Learning

Table 1

Comparison with other methods on PASCAL VOC 2012 under different partition protocols. The segmentation network is DeepLabv3+. SupOnly represents supervised training, using only labeled data.

Method	ResNet-50
	1/16 (662)	1/8 (1323)	1/4 (2646)	1/2 (5291)
SupOnly	64.20	68.30	71.87	73.12
CCT [13]	65.22	70.87	73.43	74.75
GCT [35]	64.05	70.47	73.45	75.20
CAC [34]	70.10	72.40	74.00	76.50
CPS [15]	71.98	73.67	74.90	76.15
Ours	72.48	75.10	76.10	76.40
Method	ResNet-101
	1/16 (662)	1/8 (1323)	1/4 (2646)	1/2 (5291)
SupOnly	66.87	71.54	73.69	74.22
CCT [13]	67.94	73.00	76.17	77.56
GCT [35]	69.77	73.30	75.25	77.14
CAC [34]	72.40	74.60	76.30	78.20
CPS [15]	74.48	76.44	77.68	78.64
Ours	75.32	77.80	78.95	79.16

The meaning of the bold values represent the best results.