[Retracted] Rethinking Separable Convolutional Encoders for End-to-End Semantic Image Segmentation

Table 5

Comparison of experimental results (Cityscapes validation set).

Method	Mean IoU	Mean ACC	Overall ACC
FCN-8S	0.4027	0.6241	0.6251
FCN-32S	0.3816	0.6373	0.6374
SegNet	0.4123	0.6570	0.6459
SegNet + confrontational learning	0.4275	0.6801	0.6581
SegProNet	0.4531	0.6931	0.6742
SegProNet + confrontational learning	0.4681	0.7044	0.6899
Improvement of SegProNet	0.5037	0.7247	0.7021
Improvement of SegProNet + confrontational learning	0.5219	0.7438	0.7208