Cross-domain image translation with a novel style-guided diversity loss design

作者：

Highlights：

• We design a style-guided image-to-image translation to improve the image diversity.

• A novel diversity regularizer, style-guided diversity loss (SD loss), is proposed.

• SD loss encourages the model to capture diverse image styles using style features;

• SD loss is suitable for two-domain image translation and multi-domain translation.

摘要

•We design a style-guided image-to-image translation to improve the image diversity.•A novel diversity regularizer, style-guided diversity loss (SD loss), is proposed.•SD loss encourages the model to capture diverse image styles using style features;•SD loss is suitable for two-domain image translation and multi-domain translation.

论文关键词：Cross-domain image-to-image (I2I) translation,Extracted style features,Generative Adversarial Networks (GAN),Multimodal,Multiple-domain

论文评审过程：Received 26 January 2022, Revised 15 August 2022, Accepted 16 August 2022, Available online 22 August 2022, Version of Record 2 September 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.109731