Double anchor embedding for accurate multi-person 2D pose estimation

作者：

Highlights：

• The human pose representation based on double anchor embedding can reduce the modeling difficulty of detection task.

• The parallel joint matching algorithm can improve both the accuracy and speed of joint matching.

• The results on two datasets demonstrate the effectiveness and potential of the approach in complex scenarios.

摘要

•The human pose representation based on double anchor embedding can reduce the modeling difficulty of detection task.•The parallel joint matching algorithm can improve both the accuracy and speed of joint matching.•The results on two datasets demonstrate the effectiveness and potential of the approach in complex scenarios.

论文关键词：Double anchor embedding,Ghost joint suppression,Greedy joint inference,Multi-person pose estimation

论文评审过程：Received 30 December 2020, Revised 7 April 2021, Accepted 2 May 2021, Available online 4 May 2021, Version of Record 12 May 2021.

论文官网地址：https://doi.org/10.1016/j.imavis.2021.104198