Person Re-ID through unsupervised hypergraph rank selection and fusion

作者：

Highlights：

• A novel method for fully unsupervised selection and fusion of rankers is proposed.

• The method is capable of selecting combinations in a very large search space.

• It uses hypergraph structures to encode multiple features from different rankers.

• A performance prediction measure is proposed to estimate the effectiveness of rankers.

• Very significant results were achieved for person Re-ID tasks.

摘要

•A novel method for fully unsupervised selection and fusion of rankers is proposed.•The method is capable of selecting combinations in a very large search space.•It uses hypergraph structures to encode multiple features from different rankers.•A performance prediction measure is proposed to estimate the effectiveness of rankers.•Very significant results were achieved for person Re-ID tasks.

论文关键词：Person Re-ID,Unsupervised,Hypergraph,Rank,Selection,Fusion

论文评审过程：Received 31 January 2021, Revised 3 May 2022, Accepted 8 May 2022, Available online 13 May 2022, Version of Record 31 May 2022.

论文官网地址：https://doi.org/10.1016/j.imavis.2022.104473