Prediction of partially observed human activity based on pre-trained deep representation

作者：

Highlights：

• We propose a novel ongoing human activity descriptor for video class prediction.

• The proposed method automatically captures the potentially important relationships.

• We utilize pre-trained CNN without additional network training for cost-efficiency.

摘要

•We propose a novel ongoing human activity descriptor for video class prediction.•The proposed method automatically captures the potentially important relationships.•We utilize pre-trained CNN without additional network training for cost-efficiency.

论文关键词：Pre-trained CNN,Human activity prediction,Human interaction,Sub-volume co-occurrence matrix

论文评审过程：Received 26 March 2018, Revised 3 July 2018, Accepted 11 August 2018, Available online 14 August 2018, Version of Record 27 August 2018.

论文官网地址：https://doi.org/10.1016/j.patcog.2018.08.006