Endobronchial intervention is increasingly used asa minimal invasive means of lung intervention. Vision-basedlocalization approaches are often sensitive to image artifacts inbronchoscopic videos. In this paper, a robust navigation systembased on a context-aware depth recovery approach for monocularvideo images is presented. To handle the artifacts, a conditionalgenerative adversarial learning framework is proposed for re-liable depth recovery. The accuracy of depth estimation andcamera localization is validated on anin vivodataset. Bothquantitative and qualitative results demonstrate that the depthrecovered with the proposed method preserves better structuralinformation of airway lumens in the presence of image artifacts,and the improved camera localization accuracy demonstrates itsclinical potential for bronchoscopic navigation.