Local fusion networks with chained residual pooling for video action recognition

作者：

Highlights：

• A novel multi-stage local fusion network for video action recognition

• It enhances per-frame representation by considering neighboring frames.

• The model allows end-to-end training.

• Competitive results are reported on HMDB51 and UCF101.

摘要

•A novel multi-stage local fusion network for video action recognition•It enhances per-frame representation by considering neighboring frames.•The model allows end-to-end training.•Competitive results are reported on HMDB51 and UCF101.

论文关键词：Action recognition,Residual connection,Local fusion,Deep convolutional network

论文评审过程：Received 21 April 2018, Revised 26 November 2018, Accepted 18 December 2018, Available online 28 December 2018, Version of Record 15 January 2019.

论文官网地址：https://doi.org/10.1016/j.imavis.2018.12.002