Towards a large-scale fused and labeled dataset of human pose while interacting with robots in shared urban areas