基于深度学习的图像语义分割方法研究综述

Survey of Image Semantic Segmentation Algorithms Based on Deep Learning

随着深度学习的广泛应用,语义分割的性能得到了极大的提高。总结了基于深度学习图像语义分割的研究现状,并对经典方法进行了分类梳理。首先根据数据标注类型和网络学习方式的不同,将目前主流的方法分成三类:基于全监督学习的方法、基于弱监督学习的方法和基于自监督学习的方法,并对具有代表性算法的核心思想进行了简要介绍。然后,对每类方法的性能进行了对比分析。最后,指出了图像语义分割算法存在的问题和未来的发展趋势。

With the widespread application of deep learning,the performance of semantic segmentation has been greatly improved.The research sta⁃tus of image semantic segmentation based on deep learning is summarized,and the classical methods are sorted out.First,according to the different types of data annotations and network learning methods,the current mainstream methods are divided into three categories:meth⁃ods based on fully supervised learning,methods based on weakly supervised learning,and methods based on self-supervised learning,and the core of representative algorithms The idea is briefly introduced.Then,the performance of each method is compared and analyzed.Final⁃ly,the problems and future development trends of image semantic segmentation algorithms are pointed out.