Perturbed and Strict Mean Teachers for Semi-supervised Semantic Segmentation

Benchmark Model Rank Results
semi-supervised-semantic-segmentation-on-1PS-MT (DeepLab v3+ with ImageNet-pretrained ResNet-50, single scale inference)#15Validation mIoU: 78.38%
semi-supervised-semantic-segmentation-on-10PS-MT (DeepLab v3+ with ImageNet-pretrained ResNet-101, single scale inference)#11Validation mIoU: 80.01
semi-supervised-semantic-segmentation-on-10PS-MT (DeepLab v3+ with ImageNet-pretrained ResNet-50, single scale inference)#14Validation mIoU: 78.08
semi-supervised-semantic-segmentation-on-15PS-MT (DeepLab v3+ with ImageNet-pretrained ResNet-101, single scale inference)#5Validation mIoU: 79.76%
semi-supervised-semantic-segmentation-on-2PS-MT (DeepLab v3+ with ImageNet-pretrained ResNet50, single scale inference)#13Validation mIoU: 77.12%
semi-supervised-semantic-segmentation-on-4PS-MT#11Validation mIoU: 78.20%
semi-supervised-semantic-segmentation-on-4PS-MT#18Validation mIoU: 75.70%
semi-supervised-semantic-segmentation-on-8PS-MT (DeepLab v3+ with ImageNet-pretrained ResNet-50, single scale inference)#11Validation mIoU: 79.22%
semi-supervised-semantic-segmentation-on-9PS-MT (DeepLab v3+ with ImageNet-pretrained ResNet-101, single scale inference)#12Validation mIoU: 78.72