Late feature supplement network for early action prediction

作者：

Highlights：

• A feature transfer strategy is defined to predict late features from early features.

• A new end-to-end late feature supplement-based early action prediction network is built in this work.

• The proposed feature transfer strategy can significantly improve the accuracy of early action prediction.

• The proposed method has better performance than the existing methods.

摘要

•A feature transfer strategy is defined to predict late features from early features.•A new end-to-end late feature supplement-based early action prediction network is built in this work.•The proposed feature transfer strategy can significantly improve the accuracy of early action prediction.•The proposed method has better performance than the existing methods.

论文关键词：Early action prediction,Early clip,Late clip,Feature supplement,Observation ratio

论文评审过程：Received 20 October 2021, Revised 9 February 2022, Accepted 14 July 2022, Available online 18 July 2022, Version of Record 26 July 2022.

论文官网地址：https://doi.org/10.1016/j.imavis.2022.104519