Deep Image Harmonization by Bridging the Reality Gap
This work addresses the dataset scalability issue for researchers in image harmonization, but it is incremental as it builds on existing methods.
The paper tackles the labor-intensive dataset construction problem in image harmonization by proposing a method to augment real-world datasets with rendered images and a cross-domain network to bridge the domain gap, achieving effectiveness in experiments.
Image harmonization has been significantly advanced with large-scale harmonization dataset. However, the current way to build dataset is still labor-intensive, which adversely affects the extendability of dataset. To address this problem, we propose to construct rendered harmonization dataset with fewer human efforts to augment the existing real-world dataset. To leverage both real-world images and rendered images, we propose a cross-domain harmonization network to bridge the domain gap between two domains. Moreover, we also employ well-designed style classifiers and losses to facilitate cross-domain knowledge transfer. Extensive experiments demonstrate the potential of using rendered images for image harmonization and the effectiveness of our proposed network.