CyEDA : CYCLE OBJECT EDGE CONSISTENCY DOMAIN ADAPTATION

Paper in proceeding, 2022

Despite the advent of domain adaptation methods, most of them still struggle in preserving the instance level details of images when performing global level translation. While there are instance level translation methods that can retain the instance level details well, most of them require either pre-train
object detection/segmentation network and annotation labels. In this work, we propose a novel method namely CyEDA to perform global level domain adaptation that taking care of image contents without any pre-train networks integration or annotation labels. That is, we introduce masking and cycle-object edge consistency loss which exploit the preservation of image objects. We show that our approach is able to outperform other SOTAs in terms of image quality and FID score in both BDD100K and GTA datasets.