Order-aware convolutional pooling for video based action recognition

作者：

Highlights：

• We propose a novel temporal pooling approach to aggregate the frame-level features of a video, which explores the importance of incorporating the temporal order information.

• We propose to treat the temporal evolution of the feature value at each feature dimension as a 1D signal and learn a unique convolutional filter bank for each 1D signal.

• The proposed pooling method achieves promising action recognition performance while maintaining a tractable amount of model parameters.

摘要

•We propose a novel temporal pooling approach to aggregate the frame-level features of a video, which explores the importance of incorporating the temporal order information.•We propose to treat the temporal evolution of the feature value at each feature dimension as a 1D signal and learn a unique convolutional filter bank for each 1D signal.•The proposed pooling method achieves promising action recognition performance while maintaining a tractable amount of model parameters.

论文关键词：Action recognition,Convolutional neural network,Temporal pooling

论文评审过程：Received 11 September 2016, Revised 30 January 2019, Accepted 1 March 2019, Available online 6 March 2019, Version of Record 15 March 2019.

论文官网地址：https://doi.org/10.1016/j.patcog.2019.03.002