This paper addresses the challenge of domain adaptation in semantic segmentation, where models trained on one domain may not generalize well to unseen images from different domains. The authors propose an adversarial learning method that adapts source ground truth labels to the target domain by treating semantic segmentations as structured outputs with spatial similarities between domains. They introduce a multi-level adversarial network to perform output space domain adaptation at different feature levels, enhancing the adapted model's performance. Extensive experiments on various domain adaptation settings, including synthetic-to-real and cross-city scenarios, demonstrate that the proposed method outperforms state-of-the-art methods in terms of accuracy and visual quality. The contributions of the work include a novel domain adaptation method for pixel-level semantic segmentation, the demonstration of effective alignment of scene layout and local context in the output space, and the development of a multi-level adversarial learning scheme.This paper addresses the challenge of domain adaptation in semantic segmentation, where models trained on one domain may not generalize well to unseen images from different domains. The authors propose an adversarial learning method that adapts source ground truth labels to the target domain by treating semantic segmentations as structured outputs with spatial similarities between domains. They introduce a multi-level adversarial network to perform output space domain adaptation at different feature levels, enhancing the adapted model's performance. Extensive experiments on various domain adaptation settings, including synthetic-to-real and cross-city scenarios, demonstrate that the proposed method outperforms state-of-the-art methods in terms of accuracy and visual quality. The contributions of the work include a novel domain adaptation method for pixel-level semantic segmentation, the demonstration of effective alignment of scene layout and local context in the output space, and the development of a multi-level adversarial learning scheme.