SePiCo: Semantic-Guided Pixel Contrast for Domain Adaptive Semantic Segmentation

Open paper
Benchmark Model Rank Results
domain-adaptation-on-gta5-to-cityscapesSePiCo#8mIoU: 70.3
domain-adaptation-on-synthia-to-cityscapesSePiCo#8mIoU: 64.3
domain-adaptation-on-synthia-to-cityscapesSePiCo (DeepLabv2-ResNet-101)#14mIoU: 58.1
image-to-image-translation-on-gtav-toSePiCo#5mIoU: 70.3
image-to-image-translation-on-synthia-toSePiCo#4mIoU (13 classes): 71.4
semantic-segmentation-on-dark-zurichSePiCo#6mIoU: 54.2
semantic-segmentation-on-dark-zurichSePiCo (DeepLab v2 ResNet-101)#9mIoU: 45.4
semantic-segmentation-on-gtav-to-cityscapes-1SePiCo#4mIoU: 70.3
semantic-segmentation-on-synthia-toSePiCo#4Mean IoU: 64.3
synthetic-to-real-translation-on-gtav-toSePiCo#7mIoU: 70.3
synthetic-to-real-translation-on-gtav-toSePiCo - DeepLabv2#17mIoU: 61.0
synthetic-to-real-translation-on-synthia-to-1SePiCo#6MIoU (16 classes): 64.3MIoU (13 classes): 71.4
synthetic-to-real-translation-on-synthia-to-1SePiCo (ResNet-101)#12MIoU (16 classes): 58.1MIoU (13 classes): 66.5
unsupervised-domain-adaptation-on-gtav-toSePiCo#7mIoU: 70.3
unsupervised-domain-adaptation-on-synthia-toSePiCo#6mIoU (13 classes): 71.4
unsupervised-domain-adaptation-on-synthia-toSePiCo (DeepLabv2 ResNet-101)#11mIoU (13 classes): 66.5

Rank counts only results with a code link.