Semi-Supervised Semantic Segmentation with Cross Pseudo Supervision

Benchmark Model Rank Results
semi-supervised-semantic-segmentation-on-1CPS (DeepLab v3+ with ImageNet-pretrained ResNet-101, single scale inference)#9Validation mIoU: 79.21%
semi-supervised-semantic-segmentation-on-2CPS (DeepLab v3+ with ImageNet-pretrained ResNet-101, single scale inference)#11Validation mIoU: 77.62%
semi-supervised-semantic-segmentation-on-22CPS (DeepLab v3+ with ResNet-101)#17Validation mIoU: 69.8
semi-supervised-semantic-segmentation-on-23CPS (Range View)#8mIoU (1% Labels): 33.7mIoU (10% Labels): 50.0mIoU (20% Labels): 52.8
semi-supervised-semantic-segmentation-on-24CPS (Range View)#10mIoU (1% Labels): 36.5mIoU (10% Labels): 52.3mIoU (20% Labels): 56.3
semi-supervised-semantic-segmentation-on-25CPS (Range View)#10mIoU (1% Labels): 40.7mIoU (10% Labels): 60.8mIoU (20% Labels): 64.9
semi-supervised-semantic-segmentation-on-27CPS (DeepLab v3+ with ResNet-101)#15Validation mIoU: 64.1
semi-supervised-semantic-segmentation-on-28CPS (DeepLab v3+ with ResNet-101)#14Validation mIoU: 67.4
semi-supervised-semantic-segmentation-on-29CPS (DeepLab v3+ with ResNet-101)#14Validation mIoU: 71.7
semi-supervised-semantic-segmentation-on-30CPS (DeepLab v3+ with ResNet-101)#14Validation mIoU: 75.9
semi-supervised-semantic-segmentation-on-4CPS#14Validation mIoU: 76.44%
semi-supervised-semantic-segmentation-on-8CPS (DeepLab v3+ with ImageNet-pretrained ResNet-101, single scale inference)#7Validation mIoU: 80.21%
semi-supervised-semantic-segmentation-on-9CPS (DeepLab v3+ with ImageNet-pretrained ResNet-101, single scale inference)#15Validation mIoU: 77.68%