Computer Engineering and Applications ›› 2021, Vol. 57 ›› Issue (9): 60-67.DOI: 10.3778/j.issn.1002-8331.2012-0412

Previous Articles     Next Articles

Survey of Instance Segmentation Based on Deep Learning

LI Xiaoxiao, HU Xiaoguang, WANG Ziqiang, DU Zhuoqun   

  1. 1.School of Information and Cyber Security, People’s Public Security University of China, Beijing 100038, China
    2.School of Investigation, People’s Public Security University of China, Beijing 100038, China
  • Online:2021-05-01 Published:2021-04-29



  1. 1.中国人民公安大学 信息网络安全学院,北京 100038
    2.中国人民公安大学 侦查学院,北京 100038


Target detection determines the area and category of target objects in the detected image. Semantic segmentation implements pixel-level classification of detected images. Instance segmentation can be defined as solving the problem of target discovery and semantic segmentation at the same time, and the semantics of each target instance are determined during classification. Instance segmentation networks have important application importance in areas such as drone driving, robot gripping, and industrial screening. Aiming for a blank in the current review article based on deep learning instance segmentation, this article provides an overview of single-stage instance segmentation and double instance segmentation. The stage instance segmentation classification describes different network models, focuses on the development of network frameworks over the last two years, summarizes the characteristics of each network, and suggests future development directions.

Key words: instance segmentation, deep learning, semantic segmentation



关键词: 实例分割, 深度学习, 语义分割