SCONE-GAN: Semantic Contrastive learning-based Generative Adversarial Network for an end-to-end image translation
Authors:
Iman Abbasnejad,
Fabio Zambetta,
Flora Salim,
Timothy Wiley,
Jeffrey Chan,
Russell Gallagher,
Ehsan Abbasnejad
Abstract:
SCONE-GAN presents an end-to-end image translation, which is shown to be effective for learning to generate realistic and diverse scenery images. Most current image-to-image translation approaches are devised as two map**s: a translation from the source to target domain and another to represent its inverse. While successful in many applications, these approaches may suffer from generating trivia…
▽ More
SCONE-GAN presents an end-to-end image translation, which is shown to be effective for learning to generate realistic and diverse scenery images. Most current image-to-image translation approaches are devised as two map**s: a translation from the source to target domain and another to represent its inverse. While successful in many applications, these approaches may suffer from generating trivial solutions with limited diversity. That is because these methods learn more frequent associations rather than the scene structures. To mitigate the problem, we propose SCONE-GAN that utilises graph convolutional networks to learn the objects dependencies, maintain the image structure and preserve its semantics while transferring images into the target domain. For more realistic and diverse image generation we introduce style reference image. We enforce the model to maximize the mutual information between the style image and output. The proposed method explicitly maximizes the mutual information between the related patches, thus encouraging the generator to produce more diverse images. We validate the proposed algorithm for image-to-image translation and stylizing outdoor images. Both qualitative and quantitative results demonstrate the effectiveness of our approach on four dataset.
△ Less
Submitted 7 November, 2023;
originally announced November 2023.
