Semi-supervised semantic segmentation needs strong, varied perturbations

Consistency regularization describes a class of approaches that have yielded ground breaking results in semi-supervised classification problems. Prior work has established the cluster assumption - under which the data distribution consists of uniform class clusters of samples separated by low density regions - as important to its success... (read more)

PDF Abstract
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK RESULT BENCHMARK
Semi-Supervised Semantic Segmentation Cityscapes 100 samples labeled CutMix (DeepLab v2, ImageNet pre-trained) Validation mIoU 51.2 # 2
Semi-Supervised Semantic Segmentation Cityscapes 12.5% labeled CutMix (DeepLab v2, ImageNet pre-trained) Validation mIoU 60.34% # 2
Semi-Supervised Semantic Segmentation Cityscapes 25% labeled CutMix (DeepLab v2, ImageNet pre-trained) Validation mIoU 63.87% # 1
Semi-Supervised Semantic Segmentation Pascal VOC 2012 12.5% labeled CutMix (DeepLab v3+ ImageNet pre-trained) Validation mIoU 72.45% # 1
Semi-Supervised Semantic Segmentation Pascal VOC 2012 12.5% labeled CutMix (DeepLab v2 ImageNet pre-trained) Validation mIoU 67.6% # 4
Semi-Supervised Semantic Segmentation Pascal VOC 2012 1% labeled CutMix (DeepLab v2 ImageNet pre-trained) Validation mIoU 53.79% # 3
Semi-Supervised Semantic Segmentation Pascal VOC 2012 1% labeled CutMix (DeepLab v3+ ImageNet pre-trained) Validation mIoU 59.52% # 1
Semi-Supervised Semantic Segmentation Pascal VOC 2012 2% labeled CutMix (DeepLab v2 ImageNet pre-trained) Validation mIoU 64.81% # 3
Semi-Supervised Semantic Segmentation Pascal VOC 2012 2% labeled CutMix (DeepLab v3+ ImageNet pre-trained) Validation mIoU 67.05% # 1
Semi-Supervised Semantic Segmentation Pascal VOC 2012 5% labeled CutMix (DeepLab v3+ ImageNet pre-trained) Validation mIoU 69.57% # 1
Semi-Supervised Semantic Segmentation Pascal VOC 2012 5% labeled CutMix (DeepLab v2 ImageNet pre-trained) Validation mIoU 66.48% # 4

Methods used in the Paper