Robust human motion estimation using bidirectional motion prior model and spatiotemporal progressive motion optimization

Pengle Jin, Xinguo Liu

Abstract: Highlights•We propose a bidirectional motion prior model (BMP) that generates motion priors with an optimal initial state.•We design a spatiotemporal progressive motion optimization (SPMO) method that gradually guides the optimization process towards a feasible and optimal solution.•By combining the above methods, we achieve plausible qualitative and quantitative results on the AMASS and i3DB datasets.