Monocular pedestrian orientation estimation based on deep 2D-3D feedforward

作者：

Highlights：

• 2D and 3D dimensions inputs increase accuracy of pedestrian orientation estimations

• Reuse of intermediates provides additional information of objects to network models

• Reuse of intermediates enhances logicality and interpretability of network models

摘要

•2D and 3D dimensions inputs increase accuracy of pedestrian orientation estimations•Reuse of intermediates provides additional information of objects to network models•Reuse of intermediates enhances logicality and interpretability of network models

论文关键词：Information feedforward,Logic relationship,Monocular vision,Orientation estimation

论文评审过程：Received 22 September 2019, Revised 23 November 2019, Accepted 17 December 2019, Available online 29 December 2019, Version of Record 7 January 2020.

论文官网地址：https://doi.org/10.1016/j.patcog.2019.107182