Cross-View Image Translation Based on Local and Global Information Guidance

The cross-view image translation task is aimed at generating scene images from arbitrary views. However, due to the great differences in shapes and contents of various views, quality generated degraded. Small objects, such as vehicles’ details, are not clear, which causes them be structurally inconsistent with semantic map used guide generation process. To solve this problem, we propose a novel generative adversarial network based on local global information processing module (LAGGAN) recover image’s details structures. will further combine input viewpoint target segmentation another viewpoint. proposed LAGGAN includes two-stage generator parameter-sharing discriminator. uses new (LAG) generate high-quality Moreover, integrate dilated convolutions into discriminator capture context, can enhance discriminative ability adjust LAG module. Therefore, most preserved, translated more sharply. Quantitative qualitative evaluation both CVUSA Dayton datasets attest fact that our method, LAGGAN, presents satisfactory perceptual results comparable state-of-the-art methods task.