Liu J.-LTsai AFuh C.-SHuang F.CHIOU-SHANN FUH2021-09-022021-09-02202118650929https://www.scopus.com/inward/record.uri?eid=2-s2.0-85104873031&doi=10.1007%2f978-3-030-72073-5_27&partnerID=40&md5=ef181a5887c9b6c581584f549eed89bbhttps://scholars.lib.ntu.edu.tw/handle/123456789/581300Environment semantic maps provide essential information for autonomous vehicles to navigate in complex road scenarios. In this paper, an adversarial network to complement the conventional encoder-decoder semantic segmentation network is introduced. A newly proposed adversarial discriminator is piggybacked to the segmentation network, which is used to improve the spatial continuity and label consistency in a scene without explicitly specifying the contextual relationships. The segmentation network itself serves as a generator to produce an initial segmentation map (pixel-wise labels). The discriminator then takes the labels and compare them with the ground truth data to further update the generator in order to enhance the accuracy of the labeling result. Quantitative evaluations were conducted which show significant improvement on spatial continuity. ? 2021, Springer Nature Switzerland AG.Semantics; Adversarial networks; Autonomous driving; Contextual relationships; Ground truth data; Initial segmentation; Quantitative evaluation; Semantic segmentation; Spatial continuity; Autonomous vehiclesconference paper10.1007/978-3-030-72073-5_272-s2.0-85104873031