The thesis proposes novel single-image test-time adaptation methods for image segmentation based on a deep neural network. Test-time adaptation improves the robustness of the deep neural network to data shift using a single ...