Consistency Regularization for Domain Adaptation

Collection of real world annotations for training semantic segmentation models is an expensive process. Unsupervised domain adaptation (UDA) tries to solve this problem by studying how more accessible data such as synthetic can be used train and adapt images without requiring their annotations. Recent UDA methods applies self-learning on pixel-wise classification loss using a student teacher network. In paper, we propose the addition consistency regularization term semi-supervised modelling inter-pixel relationship between elements in networks’ output. We demonstrate effectiveness proposed applying it state-of-the-art DAFormer framework improving mIoU19 performance GTA5 Cityscapes benchmark 0.8 mIou16 SYNTHIA 1.2.