Learning to Avoid Errors in GANs by Manipulating Input Spaces
This addresses a common issue in GANs for image generation, offering a simple, low-cost solution that is incremental in nature.
The paper tackles the problem of visual artifacts in GAN-generated images by manipulating the input space to avoid errors, resulting in significantly fewer artifacts with minimal impact on diversity, computation, and memory.
Despite recent advances, large scale visual artifacts are still a common occurrence in images generated by GANs. Previous work has focused on improving the generator's capability to accurately imitate the data distribution $p_{data}$. In this paper, we instead explore methods that enable GANs to actively avoid errors by manipulating the input space. The core idea is to apply small changes to each noise vector in order to shift them away from areas in the input space that tend to result in errors. We derive three different architectures from that idea. The main one of these consists of a simple residual module that leads to significantly less visual artifacts, while only slightly decreasing diversity. The module is trivial to add to existing GANs and costs almost zero computation and memory.