Review on Generative Adversarial Network in Computer Vision: Methods and Metrics
Shuai Bao · 2021
Generate adversarial networks (GAN) is a popular method, which can be widely used in numerous research areas, such as computer vision, natural language processing, and time series synthesis. However, few references are proposed to give a comprehensive review on GAN based methods. This paper aims to provide a detailed review of GAN based algorithms on computer vision tasks, such as image style transfer, image/video generation, image matting, and image super-resolution. Furthermore, we conclude the evaluation metrics for those tasks to show the effectiveness of GAN based on method. Our review can help beginners recognize the GAN based methods and give a brief introduction on how to apply them to their tasks.