Estimating 3D trajectories from 2D projections via disjunctive factored four-way conditional restricted Boltzmann machines
作者：
Highlights：
• Estimation of 3D trajectories from their 2D projections given by one camera source.
• Disjunctive factored four way conditional restricted Boltzmann machine (DFFW-CRBM).
• DFFW-CRBMs perform estimation, recognition, and future prediction of 3D trajectories.
• DFFW-CRBMs perform well while require limited amount of labeled data.
• Experiments on complex ball trajectories and human activities.
摘要
•Estimation of 3D trajectories from their 2D projections given by one camera source.•Disjunctive factored four way conditional restricted Boltzmann machine (DFFW-CRBM).•DFFW-CRBMs perform estimation, recognition, and future prediction of 3D trajectories.•DFFW-CRBMs perform well while require limited amount of labeled data.•Experiments on complex ball trajectories and human activities.
论文关键词：Deep learning,Restricted Boltzmann machines,3D trajectories estimation,Activity recognition
论文评审过程：Received 15 December 2016, Accepted 15 April 2017, Available online 29 April 2017, Version of Record 18 May 2017.
论文官网地址：https://doi.org/10.1016/j.patcog.2017.04.017
