Learning an augmentation strategy for sparse datasets

作者：

Highlights：

• Automatically learning augmentation strategy can be achieved with GANs.

• The density of the dataset influences the augmentation performance.

• Self-attention improves the GAN generator's synthetic image quality.

• Semantic reinforcement loss helps to improve synthetic image quality.

摘要

Highlights•Automatically learning augmentation strategy can be achieved with GANs.•The density of the dataset influences the augmentation performance.•Self-attention improves the GAN generator's synthetic image quality.•Semantic reinforcement loss helps to improve synthetic image quality.

论文关键词：GAN,Data augmentation,Semantic segmentation

论文评审过程：Received 20 July 2021, Revised 13 October 2021, Accepted 8 November 2021, Available online 16 November 2021, Version of Record 27 November 2021.

论文官网地址：https://doi.org/10.1016/j.imavis.2021.104338