GuidedMix-Net: Learning to Improve Pseudo Masks Using Labeled Images as Reference

Benchmark Model Rank Results
semi-supervised-semantic-segmentation-on-pascal-voc-2012-50GuidedMix-Net(DeepLab v2 with ResNet101, input-size: 512x512 with multi-scale and flip, ImageNet pretrained)#6Validation mIoU: 78.2%
semi-supervised-semantic-segmentation-on-pascal-voc-2012-50GuidedMix-Net(DeepLab v2 with ResNet101, ImageNet pretrained)#9Validation mIoU: 76.5%
semi-supervised-semantic-segmentation-on-cityscapes-100-samples-labeledGuidedMix-Net(DeepLab v2 with ResNet101, ImageNet pretrained)#8Validation mIoU: 56.9%
semi-supervised-semantic-segmentation-on-pascal-voc-2012-25-labeledGuidedMix-Net(DeepLab v2 with ResNet101, input-size: 512x512 with multi-scale and flip, ImageNet pretrained)#14Validation mIoU: 77.8%
semi-supervised-semantic-segmentation-on-pascal-voc-2012-25-labeledGuidedMix-Net(DeepLab v2 with ResNet101, ImageNet pretrained)#20Validation mIoU: 75.5%
semi-supervised-semantic-segmentation-on-pascal-voc-2012-12-5-labeledGuidedMix-Net#15Validation mIoU: 76.4%
semi-supervised-semantic-segmentation-on-pascal-voc-2012-12-5-labeledGuidedMix-Net#22Validation mIoU: 73.4%
semi-supervised-semantic-segmentation-on-pascal-voc-2012-1464-labelsGuidedMix-Net(DeepLab v2 with ResNet50, ImageNet pretrained)#17Validation mIoU: 73.7
semi-supervised-semantic-segmentation-on-cityscapes-50-labeledGuidedMix-Net(DeepLab v2 with ResNet101, ImageNet pretrained)#19Validation mIoU: 69.8%
semi-supervised-semantic-segmentation-on-cityscapes-25-labeledGuidedMix-Net(DeepLab v2 with ResNet101, ImageNet pretrained)#23Validation mIoU: 67.5%
semi-supervised-semantic-segmentation-on-cityscapes-12-5-labeledGuidedMix-Net(DeepLab v2 with ResNet101, ImageNet pretrained)#24Validation mIoU: 65.8%