Towards optimal VLAD for human action recognition from still images

作者：

Highlights：

• Tackle the empty cavity issue by properly selecting reference points.

• Reveal codeword ambiguity by imbalance assignment and enhance the reliable codeword.

• Incorporate GMP into VLAD.

摘要

•Tackle the empty cavity issue by properly selecting reference points.•Reveal codeword ambiguity by imbalance assignment and enhance the reliable codeword.•Incorporate GMP into VLAD.

论文关键词：VLAD,Empty cavity,Ambiguity,Generalized max pooling,Human activity recognition

论文评审过程：Received 19 September 2015, Revised 19 January 2016, Accepted 17 March 2016, Available online 30 March 2016, Version of Record 10 November 2016.

论文官网地址：https://doi.org/10.1016/j.imavis.2016.03.002