Cross-view action recognition with small-scale datasets

作者：

Highlights：

• Methodology for cross-view action recognition in videos, targeting small-scale datasets.

• A pipeline addressing widely diverse viewpoints, including egocentric view of actions.

• A domain adaptation procedure to transfer pre-trained features to different target views.

• A thorough experimental analysis for 2 scenarios: single view and multiple-view learning.

摘要

Highlights•Methodology for cross-view action recognition in videos, targeting small-scale datasets.•A pipeline addressing widely diverse viewpoints, including egocentric view of actions.•A domain adaptation procedure to transfer pre-trained features to different target views.•A thorough experimental analysis for 2 scenarios: single view and multiple-view learning.

论文关键词：Cross-view action recognition,Pre-trained deep features,Transfer learning,Multiview action recognition

论文评审过程：Received 17 November 2021, Revised 28 January 2022, Accepted 1 February 2022, Available online 5 February 2022, Version of Record 17 February 2022.

论文官网地址：https://doi.org/10.1016/j.imavis.2022.104403