Research on Reinforcement Learning algorithms in Computer Vision

Abstract

As artificial intelligence technology continues to develop, reinforcement learning (RL) is evolving as a potent form of artificial intelligence. Reinforcement learning, as a subfield of machine learning, focuses on how to behave in a given situation in order to maximize the expected rewards. Due to the excellent perceptual and decision-making capabilities of RL algorithms, reinforcement learning has been widely used in various fields including medicine, finance, robotics, video games, and computer vision (CV). Among them, computer vision is a challenging and significant research subject in both engineering and science fields. Because diversity and imperfections are prominent features of the CV domain, there are numerous ways to utilize reinforcement learning to enhance CV tasks. This paper aims to introduce the fundamental concepts and methodology of reinforcement learning. Moreover, this paper details the recent applications of reinforcement learning in different branches of the CV field, and makes a comparison of the performance of the different algorithms involved.