Confidence-Weighted Boundary-Aware Learning for Semi-Supervised Semantic Segmentation

Benchmark Model Rank Results
semi-supervised-semantic-segmentation-on-cityscapes-100-samples-labeledCW-BASS (DeepLab v3+ with ResNet-50)#3Validation mIoU: 65.87
semi-supervised-semantic-segmentation-on-pascal-voc-2012-50-labeledCW-BASS (DeepLab v3+ with ResNet-50)#3Validation mIoU: 77.15
semi-supervised-semantic-segmentation-on-cityscapes-6-25-labeledCW-BASS (DeepLab v3+ with ResNet-50)#11Validation mIoU: 75.00
semi-supervised-semantic-segmentation-on-cityscapes-12-5-labeledCW-BASS (DeepLab v3+ with ResNet-50)#12Validation mIoU: 77.20%
semi-supervised-semantic-segmentation-on-cityscapes-25-labeledCW-BASS (DeepLab v3+ with ResNet-50)#13Validation mIoU: 78.43%
semi-supervised-semantic-segmentation-on-pascal-voc-2012-92-labeledCW-BASS (DeepLab v3+ with ResNet-50)#14Validation mIoU: 72.8
semi-supervised-semantic-segmentation-on-pascal-voc-2012-12-5-labeledCW-BASS (DeepLab v3+ with ResNet-50)#17Validation mIoU: 75.81%
semi-supervised-semantic-segmentation-on-pascal-voc-2012-25-labeledCW-BASS (DeepLab v3+ with ResNet-50)#19Validation mIoU: 76.2