Semi-supervised Semantic Segmentation with Prototype-based Consistency Regularization

Benchmark Model Rank Results
semi-supervised-semantic-segmentation-on-pascal-voc-2012-50PCR (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K)#2Validation mIoU: 80.91%
semi-supervised-semantic-segmentation-on-pascal-voc-2012-25-labeledPCR (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K)#7Validation mIoU: 80.78
semi-supervised-semantic-segmentation-on-pascal-voc-2012-12-5-labeledPCR (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K)#8Validation mIoU: 80.71%
semi-supervised-semantic-segmentation-on-pascal-voc-2012-6-25-labeledPCR (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K)#9Validation mIoU: 78.6
semi-supervised-semantic-segmentation-on-cityscapes-6-25-labeledPCR (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K)#13Validation mIoU: 73.41%
semi-supervised-semantic-segmentation-on-cityscapes-25-labeledPCR (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K)#14Validation mIoU: 78.4%
semi-supervised-semantic-segmentation-on-cityscapes-50-labeledPCR (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K)#14Validation mIoU: 79.11%
semi-supervised-semantic-segmentation-on-cityscapes-12-5-labeledPCR (DeepLab v3+ with ResNet-101 pretraind on ImageNet-1K)#16Validation mIoU: 76.31%