ProCST: Boosting Semantic Segmentation Using Progressive Cyclic Style-Transfer

Benchmark Model Rank Results
domain-adaptation-on-gta5-to-cityscapesDAFormer + ProCST#9mIoU: 69.4
domain-adaptation-on-synthia-to-cityscapesDAFormer + ProCST#10mIoU: 61.6
image-to-image-translation-on-gtav-toDAFormer + ProCST#7mIoU: 69.4
image-to-image-translation-on-synthia-toDAFormer + ProCST#6mIoU (13 classes): 68.2
semantic-segmentation-on-gtav-to-cityscapes-1DAFormer + ProCST#5mIoU: 69.4
semantic-segmentation-on-synthia-toDAFormer + ProCST#5Mean IoU: 61.6
synthetic-to-real-translation-on-gtav-toDAFormer + ProCST#10mIoU: 69.4
synthetic-to-real-translation-on-synthia-to-1DAFormer + ProCST#7MIoU (16 classes): 61.6
unsupervised-domain-adaptation-on-gtav-toDAFormer + ProCST#9mIoU: 69.4
unsupervised-domain-adaptation-on-synthia-toDAFormer + ProCST#8mIoU (13 classes): 68.2